CA2116677A1 - Transaction document reader - Google Patents

Abstract

A transaction document reader reads image data on a transaction document. The image data includes marking areas employed by a user to record marks and includes characters. A sensor (52) senses, one row at a time, rows of image data as the document is transported across the sensor. Each row includes a number of pixel areas.
The pixel areas of the rows being aligned in columns. The sensor generates output signals representative of the pixel areas. The output signals are transformed into bit signals (72) representative of each pixel area. The bit signals are stored in image memory (62) which is organized into rows and columns of bit signals corresponding to the rows and columns of the pixel areas of the image data. The marking areas are located in image memory. The marks are identified in the marking areas while additional rows of bit signals are being stored and while other marking areas are being located by probing the bit signals in image memory. A specific area within image memory representative of a single character is also located. A neural network is utilized to recognize the character within specific area.

Description

W0~3/05480 -1- 2 ~ 1 6 6 7 7 ~ PCT~US9~/07260 TRAN8A~TIO~ ~OC~NT READ~R
,. ., -:
Fiel~ Qf th~ Inv~io~
This invention relates to the field of optical - -readers. More particularly, ~he invention xelates to the recognition of handwritten marks and printed characters on transaction documents.
5 Ba~groun~ of the I~ve~tion ~ -Handwritten marks are often record d in boxes, marking areas, or other fields on a transaction document.
These handwritten marks take many different forms. For example, the handwritten mark may be a dot, an "X", a ch~ck 10 mark, or the handwritten mark may be a cratching out of a mark previously made by the user. Because many transaction documents, such as lottery documents, restaurant ordering documentc, standardized test taking documents, are utilized, it is neceqsary to be able to interpr~t such 15 handwritten marks thereon.
Past efforts for interpreting such marks included aligning ~he marking areas in a number of columns on a transaction document. Individual ensor~ positionally aligned with the marking areas on the document optically 20 read the marks. Nisalignment of the sensors ov~r the particular columns of marking areas produced incorrect readings.
Other optical readers are known in the art for certain uses, such as lottery documents, multiple choice 25 transaction documents, standardized test form documents, etc. U.S. Patent No. 4,724,307 to Dutton et al. discloses a marked card reader.~ The reader, as best understood, images an entire ~arked card and utilizes identification marks thereon to identify the type o~ card being read and 30 various baselinè information with regards thereto, ~uch as the po ition of image fields. The identification marks --also appear to provide a means for compensating for misalignment of the marked card as it is read by an imaging ~ -device. The compensation is provided by recalculating 35 addresses of image data in memory such that a corrected memory compensating fox the misaligned document is created.

SllBSTlTUTE S~ :

WO9~/05480 2 t ~ ~ ~ 7 7 PCT/US92/07260 A device like that dascribed in Dut~on et al.
images the en~ire document prior to reading any information on the document. The speed of reading the information on the documen~ is therefore, limited. In additiQn, the 5 recalculation of addresses for the memory containing ~he image data to correct for misalignment, reduces the speed by which ~he dorument can be read.
Dutton et al~, in performing the reading of handwritten marks on marked cards, compares ~he gray level 10 value of pixels representative o~ the data on ~he card at poten~ial mark areas with the gray level values for pixels surrounding the potential mark area to determine whether a mark is present or not present. Such gray level pixel value to gray le~el pixel value comparison consumes 15 precious time in the reading process.
An optical reader capable of both handwritten mark recognition and predetermined character recognition, such as the recogn-ition of numerals 1, 2, 3, etcO, is useful in providing a tool ~hereby later processing of ~he 20 handwritten ~arks can be analyzed according to certain recognized characters. For example, a lottery document having a numbe!r of boxes with handwritten mark~ therein may be analyzed after the marks are read according to later processing circuitry initiated by the optical recognition 25 of certain characters on the ~ame document, such as the number "99". ~he problem of time con~umption asEociated with reading handwritten marks on documents is equally applicable to the recognition of predetermined character~
on documents. In order to improve the number of 30 transactions proces~ed in a period of time, the speed and accuracy of optical readers need to be increased.
For the processing of many transaction documents, it is necessary that handwritten mark~, or the interpretation thereof, be associated with ~ignatures or 35 other in~ormation about the perqon who placed the handwritten ~ark. Such asæociation can be accompli~hed by user-carried information storage cardsl for example, those SllBSTlTUTE SH~T

W093/0~80 2 ~ ~ 6 6 7 7 PCTIUS92/072~0 known and u~sd in connection with identi~ica~ion functions and for ~anaging the debiting and crediting o~ customerc financial account~.
Europ~an patent application 0 370 925 di closes 5 programmable game entry card for use with a wagering system. The player card is provided with connection to a wagering terminal and ~he information tored in memory on the card can be accessed by the terminal. The card includes the user's status information, wagering amounts 10 information, and other applicable information. Such information is updated when a card transaction is completed. ~he card allows the player to forego 60me steps in the wagering proc~ss. User-carried information storage cards as described in EP A 0 370 925 would need to be 15 utilized on il regular basis to make them cost efficient.
. .
A simpler and less expçnsive card eystem accomplishing some of the tedious tasks of a lottery procees or transaction proces~ would be beneficial.
Transaction documents are normally printed in few 20 colors in ordler to render the printing invisible to optical readeræ while allowing handwritten marks readable. If transaction documents were printed in more colors, enhanced use of the document may be likely, due to aesthetic ~ppeal.
Furthermore, various features on document~ could be 25 separable by means of making different ~eatures appear in different colors and reading those fea~ures. For example, marking areas on a document ~ay appear in red, character~
for identifying the documents may appear in green, etc. An optical card reader which could discern features printed in 30 various colors would allow such documents to be printed in the various colors, enhancing ~lexibility of document design.
Prior systems attempting to read colored imaga data on a transaction document, spacially printed portions 35 of ~he docu~ent in one color and other portions in ano~her color. A light source associated with an optical scanner would utilize filters of various colors to achieve SlJBSTlTlJTE SHEE7~

W~93/~5480 . PCT/US92/07260 21166'~7 4 recognition o~ the various color data. For example, a green filter would be used to provide refl~ction of grQen light from portions of the document to an optical sensor and a red filter to reflect red light fro~ another portion 5 of the documen~ to read the image data prin~ed ~hereon. The problem with filtering the light is that isolation of the green and red filtered light at the boundary b~tween the red and green image data is difficult a~d result~ in inflexible designs of the document. A new technique for lO rendering colored image data "visiblel' to an image processing system would be beneficial.
Increasing the speed of optically reading handwritten marks and characters is an important characteristi.c of a successful optical reading system. A
15 system which dynamically reads both handwritten marks and charact~rs e].iminates the need for duplicate apparatuses.
An optical reader having both mark recognition and character recognition could be used to facilitate document transactions such that users could eliminate tedious steps 20 in a transac1:ion process. Printing various features in di~ferent col.ors on a transaction document and being able to accurately r~ad these features would facilitate perception and use by a user and allow flexible document production. A need exists for a transaction document 25 reader which provides one or ~ore o~ these characteristics.
~ummary.of the ~nve~tio~
The apparatus of the present inven,tion provide~
a meanC for reading i~age data on a ~ransaction document.
The i~ag~ data includes ~arking areas employed by a user to 30 record mark~ and includes a predetermined 6et of character~. Bit signals representative of rows and columns of dark or light pixel areas of the image data are g~nerated. The bit ignals are single bit signals, however, multi-bit signals can also be utilized. ~n image 35 memory stores ~he bit ignals. The image ~emory i8 organized into rows and columns of bit signals ~orresponding to the pl~rality of rows and columns of pix~l SU~ST~TUTE S~

WO 93/05480 PCltl~S92/07260 .~ ~
s 2~1~i677 areas of the image data.
The maxks recorded in the marking areas are read by lo~ating a por~ion of th~ image m~mory representative of ~he marking area~ of the tranæaction document a~ a 5 function of particular reference characteristics of the document and the marks are identified in the marking area~
by probing rows of bit signals of the image memory representative of rows of pixel areas within the marking areas.
Charac~ers are optically read by locating a specific area within a character area of the image memory which is representative o~ at least a single character. A
neural network for recognizing the predetermined set of characters i5 then applied to the specific area.
The apparatus of the present invention also dynami ally reads the marks and predetermined characters.
A ~ensor senses rows of image data a~ the document is transported across the sensor. Each row of image data include~ a plurality of pixel areas. The pixel areas are 20 aligned in a plurality of columnc. The sensor generates output ~ignalls representative of the pixel areas. Transfor~
circuitry rec:eives the output ~ig~al~ and generates bit signal~ repr~sentative of each pixel area. An image memory stores the bit signals in rows and column~ corr~sponding to 25 the plurality of rows and columns of pixel areas of the image data.
~ he marks recorded in the marking areas are read by locating a portion of the i~age m~mory repre~ntative of the marking areas of the tran~action document. The 30 portions of the i~age memory are located a~ter rows of bit signals representative of the marking area are ~tored in memory and while additional rows of bit si~nals representative of the other image data are being sensed, tran~formed, and stored in the image memory. The ~arks are 35 identified while the additional rows of bit signals are being ~tored and while the marking areas are being located by probing the bit signals within ~he image memory.

'..' '~ -'~ ~' '' W093/0~80 PCT/US92/07260 2116~77 6 Characters are optically read by locating a specific area within a character area of the image memory which is representative o~ a single character after the bi~
eignals representative of the ~ingle charac~er are stored 5 in the ima~e memory and while additional bit signal representative of additional image memory data are being sensed, tran~formed and stored in the image memory. After normalization, a neural network for recognizing the predetermined set of characters is then appiied to the 10 speci~ic area.
The apparatus of the present invention also determines a type of transaction document having image data thereon by generating bit signals representative of rows and columns o~ dark or light pixel areas of the image data.
15 The bit signals are stored in an image memory organized into rows and columns corresponding to the pixel areas o~
the image data of the document. The document is identified as a function of the dimensions of the document after probing at least one row of the image memory to determine 20 the document~ dimensions. In one embodiment, the document's w.idth determines the type of transaction document.
The invention further includes apparatus ~or compensating for distortion caused by a linear array sensor 25 and optics. The linear array sensor senses a plurality of rows of image data of a document transported across the ~ensor. Each row of image data includes a plurality o~
dark or light pixel areas. The pixel areas are aligned in ~ plurality of colu~ns. Bit signals representative of the 30 dark or light pixel areas ~re generated and ~tored in an image memory organized into rows and colu~ns of bit signals corresponding to ~he rows and columns of pixel areas of ~he image data on the document. The compensation for distortion caused by the linear array sensor and optics is 35 a~complished by calibrating the document reader using a master document. The master document, having lines corresponding to the transaction document, is transported SuBs~7rvrE S~EE~

W093t0~80 PCT/VSg2/07260 7 21 1 ~ 77 across the sensor. By comparing the pr~determined position of ~he master document lines measured wi~h respect to ~::
document reference characteristics prior to tran~porting the master document across the ~en~or, to the line as 5 represented by the bit signals in the image memory after the ma~ter document is sensed and bit ~ignals are generated, the compensation required is determined. As read areas on the transaction document are being located and read by probing the bit signals in ~he image memory, lo the probing is adjusted to probe the correct bit signals within the read area in response to the comparison.
The apparatus of the present invention al~o includes appa:ratuæ for compensating for mi~alignment of the document as it is transported across the sensor. Bit 15 signals are generated as discussed above with regard to compensation for dis~ortion by the sensor. To compensate for misalignment, document r~ference characteristics are located with-in the image memory and the angle of the document is determined with respect to a line established 20 by the linealr array ~ensor. When a read area of a transaction document is being located and read by probing the bit ~ignal~ of the image memory, the probing is ~ -adjusted to probe correct pixel areas within the read as a function of the angle of the document. ~ ; -The apparatus of the prasent inv~ntion is utili2ed in a m~thod for facilitating transactions involving transaction documents. T~e ucer i~ provided with a user id~ntification card having an ide~ti~ication number ~ :~
thereon and also with a user per~onalized document, The 30 user identification card ~u~ber and the user per onalized ! ~ ,:; . ' document are digitally imaged by ~he reader. The digital .
information representing ~he user identification number and the associated user personalized document is stored. A
document transaction can then be completed by transporting 35 a transaction document and the.uBer identification card with the user identification card number ~hereon through :~
the reader. The identification number is optically , SVE;STITU~E ~YE~

W093/0~80 PCT/US92/07?~

2 ~ 7 7 8 recognized and the transaction document read. The identification number is associated with the tran~action document, whereby a personalized document is not required with each tran~action.
The present invention also includes apparatus for reading colored image data on a document. A sensor sen~es a plurality of pixel areas o~ image data on the document and generates output signals representative of th~ pixel areas. The output signals are transformed to digital 10 signals representative of the pixel areas and stored in image memory. The image memory include~ at least twc color image memory fields for different colors. The color image memory fields are probed to read image data of a particular color.
In one embodiment of reading colored image data, the sensor includes a first color light source and a second color light source. The first and second color light source alternately illuminate a plurality of row~ of pixel areas of the document. The sensor ~enses reflected light 20 from the document and generates output ~ignals representative of the pixel area~.
In another embodiment of reading colored image data, the sensor include~ a full or broad spectrum light source and a color sensor for sensing reflected light of 25 different color~
Brief Dss~ri~t~on o~ ~he Dra~n~
Thu~ly summarized, the present invention ~ay be ~etter understood, and ~he advantages made apparent to those skilled in the art, by reference to the accompanying 30 drawing wherein like reference nu~bers refer to like ! ~ -;.
element~ in ~he ~everal figures and in which:
- Fig. 1 i~ - a perspective view of a transaction document reader;
Fig. 2 is a side schematic view of ~he transaction document reader of Fig. 1;
Fig. 3 is a top schematic view of the transaction document reader of Fig. 2;

SUB~ S~lEET

W0~3/0~80 PCT/USg2/072~
9 2~1~677 Fig. 4 is a scAematic diagram of the reader circuitry of the transaction document reader o~ Fig~
2;
Fig. 5 is a representative diagram of a transaction document to be read by said transaction document reader o~ Fig. 2;
Fig. 6 i~ a r~presentative diagra~ of a ::
receipt to be read by tran~action document reader of Fig. 2;
Fig. 7 is an enlarged view of a portion of the transaction document in Fig. 6;
Fig. 8 is an illustration of an image memory represen$ative of the enlarged view of Fig. 6; :
Fig. 9 is a master calibration document for calibrating the transaction document reader of Fig. 2 with re~pect to the document o~ Fig. 5; ~ :
. Fig. 10 is an enlarged view of a portion of the transaction document of Fig. 5; .i ~ ii.
Fig. 11 is an illustration o~ the image memory repres~ntative of the portion of Fig. 10; :.:
Fig. 12 is a schematic representation of a normaliz~d Fig. 11;
Fig. 13 iB an additional embodiment o* a i~
transaction document to be read by the transaction document reader of Fig. 2; ~i Fig. 14 is an enlarged view of a portion Qf the.transaction document of Fig. 13;
F~g. 15 is an illu tration of an image .
memory representatiYe of the portion of the . tran action do~ument sho~n in Fig. 14;
Fig. 16 is an illustration of a ma ter calibration document used to calibrate the transaction :~ :~
document reader of Fig. 2 with respect to the document o~ Fig. 13; :
Fiq. 17 is a schematic illustration for compensating for misalignment of a transaction document as it is transported and raad by the :: :~., . ::

-~ -W093/05480 2~6677 PST/US92/072~
, "
transaction document reader of Fig. 2;
Fig. 18A, 18B, and 18C are schematic representations of distortions requiring compe~sation;
Fig. 19 i~ a ~chematic illustration o~ a Smarking area of a transaction document read by a transaction document reader of Fig. 2 wherein said light ~ource is an alternating light source;
Fig. 20~, 20B, and 20C is a sche~atic representation of an image me~ory representative of a marking area when said light source is an alternating color light source;
Fig. 21 is a portion of a transaction ~ :
document, such as that of Fi~. 5, printed in two different colors; :::
Fig. 22 is a schematic illustration of an ; ~ :
image m,~mory representative of the portion of the :
transact:ion document shown in Fig. 21;
Fig. 23A, 23B, 23C, and 23D, are views of - :
cards to be used in a transaction method utilizing the ~ransaction document reader of Fig. 2;
Fig. 24 is a flow diagram of the reading of image data o~ the present invention;
Fig. 25 is a flow diagram of the calibration of the transaction document reader;
Fig. 26 is a flow diagram of the xeading of ~
marking areas of Fig. 24; ~ :
Fig. 27 is a flow diagram of the optical reading of characters of Fig. 24; and Fig. 28 shows a block diagram of an alterna~ive embodiment of a portion of Fig. 4.
D~taile~ P~oriptio~ of_ th~ Pr~rre~ ~mbO~m~8 ~ In the following description o~ the pre~erred embodiments, reference i~ made to the accompanying drawings which fo~ a part hereof, and which show the preferred 35 embodiment It is to be understood, however, that other embodiments may be utilized and ~tructural changes ~ade without departiny from the scope of the present invention.

SUBSl l~UTE SHEET

W093/0548~ PCT/US92/072~
11 ` 2116677 Fig. 1 illustrate~ a transaction document reader 20 and a transaction documant 90 placed upon a reader receiving platform 12. Reader 20 includes u~er input pad 18 and display 14 for facilitating a do~ument tr~nsaction.
5 ~he transaction document so is shown in detail in Fig. 5.
Transaction documents are used for playing the lottery, for taking tests, for ordering items, etc. The transaction document 90, Fig. 5, includes a character area 92 for printing of predetermined characters 104. Al~o printed on 10 the transaction document are marking areas 94 for a user to place a handwritten mark therein in order to complete a transaction such as choosing lottery numbers.
The document 90 i~ placed on reader receiving platform 12 and inserted into reader 20. Therein, the 15 document 90 is transported through the reader 20 by transport mec:hanism 220 The transaction document reader 20 linearly optically scans the ~ran~action doc~ment 90, sensing rows of pixels 102. The pixel areas of the rows are aligned in columns of pixels 100. The transaction 20 document reader 20 includes a linear image CCD ~ensor 52 which provides analog output ~ignals 53, represent~tive of the rows of pixel areas 102, to the reader circuit~y 54 which processes the analog 5ignal5 53 on a row by row ba~is.
The reader cirauitry 54 includes a ~otor control and sensor portion 56 for mechanically controlling the transport mechanism 22 and transaction document reader 20.
The output si~nals 53 from CCD array 52 are applied to transform circuitry 60 under control of CCD array timing 30 control 64. The analog output signals 53 are transformed to single ~it signals and ætored in image memory 62~ The i~age memory 62 is then probed or interrogated on a dynamic basis by proce~ing and control circuitry 58 under the control of ~oftware 68. The single bit signals stored in 35 image memory 62 are probed to recognize whether mark~ g6 exist on document 90, Fig. S, and to recognize the predetermined characters 104 in ~he character area 92 of - -: ~ .
SUBSTITUTF SHF~

W093/0~80 2 1 1 6 6 7 7 PCT/U~92/07260 document 9o. It would be readily apparent to one skilled in the art that instead of utilizing single bit signals, that ~ultiple bit signals can be u~ed to implement the transaction document reader ystem.
With reference to Figs. 2-12 and Figs. 24-27, the reading of tran~action document so, Fig. 5~ by transaction document reader 20 will be described in further detail.
Fi~. 24 gener~lly describes the flow of the syste~
portions of which are under the control of software 68.
10 Each of the steps generally described in Fig. 24 shall be described in further detail below and with re~erence to . Fig. 25-27.
Transactisn document reader 20 includes transport mechanism 22, sensing device 34 and reader circuitry 54.
~5 Transaction document 90 is placed, printed material face down, on readler receiving platform 12. As the document is placed over :Leading edge sensors 26 and rail ~dge side sensors 24 o~E the transport mechanism 22, Fig. 2, bias drive roller 23 is activated by motor controller 56, Fig.
20 4 and document transport is started, Fig. 24. Bias drive roller 28 is positioned at an angle with respect to ~he side rail 29, Fig. 3, and drives the document towards the side rail such that the documents rail side edge 106, Fig.
5, is in contact with side rail 29. The bias drive roller 25 28 keeps ~he document aligned by keeping the rail side edge of the document orthogonal with the linear scan as the document is transported acros~ window 42 to be sen~ed by sensing device 34.
Sensing device 34 includes window 42, reflective 30 edge window ~ensor 36, light sources 38, fixed ~irror 46, lenses 48 and 50, and CC~ linear image censor 52. As document 90 i~ transported by bias drive roller 28 2cross window 42, the leading edge 104 of document 90 triggers reflective edge window ~ensor 36 ~uch that light sources 38 35 are activated to illuminate the document as it is transported across the window 42. By sensing when the document is coming upon the window 42 by means.of SUBYITUT~ ~3HEt W093/0~80 13 2 1 ~ ~ ~ 7 7 PCT/VS92/072~

reflective edge window ~ensor 36, and initializing illumina~ion accordingly, the time required for light sources 38 to be activated i8 minimized. Drive rollers 30 are al50 ~c~iv~ed by the sensor~ 24, 26 and keep the document 90 continuously and constantly mo~ing ~cross the window 42.
As the trailing edge llO of document 9O triggers trailing edge sensor 32, the motor driving drive rollers 28 and 30 are turned off and document 9O has completed lO transport through the r~ader 20. It should be r~adily apparent that the trailing edge sensor's function can easily be accomplished by setting a predetermined time necessary ~or the motor driving the rollers 30 and 28 to be activated such that the document is completely transported 15 across the window 42 and automatically shutting off the motor after such predetermined time. The document is transported t~rough the system at a particular speed as a function of lthe required scanned resolution and document processing ti~e.
As shown in Fig. 2 and Fig. 3, document 90 i5 tran~ported alcro~s window 42 and light ~ourceæ 38 provide illumination thereof via light rays 40. Light sources 38 include Qither di~crete LED's, (Light Emitting Diodes) or LED chips on a board and illuminate the document 9O with a 25 cingle color of light, normally red or green. Reflect~d light 44 i8 reflected from document 90 through window 42 and i8 the~ reflected from a front surface mirror ~6. The reflected light 44 is foaused by i~age ~aanning lens 48 onto linear i~age CCD sensor 52 ~hrough field flattening 30 lens 50. The image scanning lens 48 is a 12.5 mm lens, F2.9, manu~actured by Universe Kogaku ~America), Inc., part number GT-12M. CCD llnear imag~ 8ensor s2 is a sensor such as a LORAL Fairchild Imaging Sensor CCD134 or a Toshiba CCD
linear i~age sensor TCD-132D, 1024-element high-speed 35 linea~ image sensors.
The linear sensor 52 produces an analog signal for 1024 pixel areas across a row of pixel areas 102 of the SUB~TITUTE SHEE~

W093/05480 2 1 ~ 6 ~ 7 7 PCT/US92/07260 document 90 as the document is being scanned. The analog signal outputs 53 are representative of the total light radiation ref lectPd from aach pixel area of the row of pixel areas. An analog shift registar i~ clock@d 5 periodically under control of CCD array timing control 64 to ~hift the 1024 analog signals and to clear the linear sensor 52 to receive a new reading of 1024 analog signals of the next row of pixel areas as the document 90 is conti~uously moved across the window 42. The sensor read~
10 a matrix o~ 1024 X approximately 1300 pixel areas for a ~ix inch transaction document. To on~ skilled in the art it should be readily apparent that the number of pixel areas read by the image ensor, ~he size of the document, and the size of the image data on the document, are easily changed.
15 The illustrative documents described herein are simplified for facilitat.ing description of the transa~tion document reader.
The analog signals 53 representative of a row of pixel areas llD2 from linear CCD sensor 52 are applied to 20 analog digita:l (A/D) converter 70 of transform circuitry 60. Transfo~ circuitry 60 includes A/D converter 70, transform memory 72, and transform control 74. Analog to digital converter 70 provides a 7-bit output signal representative of the gray scale of each distinct pixel 25 area of the transaction document 90. The 7-bit output signal and signals from transform control 74 representative of the column number of each pixel area in a row, addr~s~es the transform memory 72 to select a single bit ~ignal (one or zero) for each pixel area. The signals fr~m transform 30 control 74 are generated as a unction of timing ~ignals from CCD array timing control 64. The single bit signals are th~n stored in image memory 62. As rows o~ ~ingle ~it signals are stored partial image capture is compl~ted, Fig.
24.
Transform memory 72 is a threshold bit map of individual columns in a look-up table form. The thresholding of transform memory loads the memory with a SUBSTITUTF SHEET

W093/05480 PCT/US9~/07260 ~ .
2 1 ~ ~ ~ 7 7 , " .
map o desired thresAold results. For example, if the threshold value for minimum white for a particular colu~n i5 47 of a gray ~evel sensitivity of 128, addre~se XX47 and aboY~ for an individual column in the lookup table 5 would contain a "1" and addresses XX46 and below would contain "0". Thre holding of transform memory is completed prior to partial image capture of the tran action document ,. .. .
9 0 .
The reading of a plain reflective document, 10 normally white ~not shown), thru proce~sing of the reflective document's image in image memory 62, cau~es the transform memory 72 to be mapped with ones and zeros or single bit ~ignals. The reflective document as it is scannPd by doc:ument reader 20 identifies the maximum white .5 gray value for each of the 1024 sensor areas of each row of pixel areas by aYeraging the gray ~cale values for the pixel areas in each coluDn. The look up table is mapped by setting a predeter~ined fraction of the maxim~m white gray scale values, as read by the CCD array when the 20 reflective document is transported across window 42 and scanned by CCr) image 6ensor S2, as the threshold value ~or the column. The addresses corresponding to the ~hreshold value ~nd the addresses above the threshold value are loaded with ones and those.corresponding addresses below 25 the threshold value are loaded with zeros.
The predetermined fraction i~ typically approximately 40 to 50 percent. For example, if the averaging of gray cale values gives a maximum white of 117 in gray seale value for a partic~lar column of the row, ~he 30 threshold value is sele~ted as 40 percent of that value or 47. Therefore, addresses corresponding to values 0 to 46 are mapped ~uch ~hat a corresponding pixel area having a gray scale value of 0 to 46 i~ preæumed black or dark and a "0" bit is later read from trancform memory, and 35 addresses are mapped ~uch that a ~orresponding pixel area having a gray scale value of 47 to 127 is presumed white or light and a "1" bit is later read from transform memory 72.

SUBSTITUTE S~EET
, ;, ~

W093/0~0 PCT/US92/07260 21~6~77 16 The process of reading the plain ref lective document compensates for the lack of uniform illumination by the light ource 38 acro~s the row of pixel areas, compensates for reduc~ion of the light reflected by ~he 5 do¢ument as it is focu~ed through side portions of the lenses, and for compensates variations in CCD linear image sen or pixel cells. Reading of the plain reflecti~e document need only be done once Unle5S characteri~tics of the reader change.
A~ter the transform memory 72 is mapped a transaction document 90 is transported through the reader, and capture of the image data can be accomplished. The 7--bit gray ~cale values from A/D converter 70 and signals from transfo~ control 74 representative of the column 15 number in ~he row of pixel areas of the document scanned, look up for t:he pixel area represented by the 7-bit qray scale value a.nd column position signal whether the pixel area is tG be represented by a "1" or a "0"', i.e. whether the pixel are,a is light or dark. If the pixel area is a 20 dark pixel ar,ea, a ~ingle bit .ignal, a zero, is provided to image memory 62. If the pixel area i~ a light pixel area, a single bit signal, a one, is provided to image memory 62. The gray scale value can al80 be ~tored in image memory 62.
Image memory 62 is arranged in rows and column corresponding to the pixel areas ~ensed by the linear array sensor 52. For example, row~ of pixel area~ 102 correspond to rows o~ bits 134, Fig. 8 t and colu~ns of pixel areas corre~pond to columns o~ bits 132, Fi~. 8. Each and every 30 pixel area is u~ed to address transform me~ory 72 and results in a single bit signal, a zero representing a dark pixel area and a one representing a light pixel area, being stored in image memory 62.
The rows of pixel areas are transformed into 35 single bit æignals, on a pixel by pixel basis, and applied to image memory 62 on a pixel by pixel basi~. By continuously ~ensing pixel areas of document 90 and W093/0~0 ~ PCT/US92/0726~ ~-17 211~77 ~:;
transforming Eaid pixel areas into ~ingle bit signals on a pixel by pixel ba~is, the single b~t signals of ~ubcequently 6en~ed row6 of pixel area6 are transformed and stored in image memory 62 while a portion of the image 5 memory 62 already containing stored single bit signals are interrogated. As such, as shown in Fig. 2~, the image partially captured allows for the document to be identified and some marks and characters to be read as the capture of the rest of the document continues.
Motor controller and sensor 56, Fig. 4, receives signals from LED sensors 24, 26, and 36. In response to sensor signals, motor controller and sensor 56 controls ~he motor by signals 76 which drive the bias drive roller 28 and drive rol].er 30. Controller and sensor 56 also enable as the illumination of the document 90 via a light Qource 38 per signals orl line 80. In addition, the motor controller and sensor provide a timing signal to CCD array timing control 64. Timing control 64 facilitate~ dynamic capture of ~he transaction document as it is being sensed and 20 transformed by relaying timing information to image memory 62 and proces~;ing and control circuitry 58. Such timing signals are recognized by the circuitry and enable the circuitry under control of software 68 to begin interrogation of image memory 62 ~or recognition of 25 handwritt~n marks 96 on document 90 or for optically recognizing the chara~ters 104 thereon. For example, after approximately thirty rows of pixel areas have been sensed, transfor~ed and ~tored in image ~emory 62, ti~ing control relays information to processing and control 30 circuitry 58 to begin interrogation o~ the single b~t signals in image memory 62.
Txansaction document 90, Fig. 5, includes 8iX
marking areas 94, two of which have marXs 96 therein.
Four of these ~arking areas 94 are enlarged and illustrated 35 in Fig. 7~ Tran~action document 90 also includes character area 92 and a signature/address block 98. The edges of transaction document 90 include rail side edge 106, ~Bgm~E S~

W093/054~0 ~1 1 6 6 7 7 PCT/US~2/07260 opposite edge ~08, leading edge lo~ and trailing edge llo.
Prior to scanning and rsading transaction document 90 and af~er thresholding the transform ~emory ~2, a master calibration documen~ 150, Fig. 9, directly S corresponding to ~ransaction document ~O i5 transported ~hrough the reader 20 and scanned by linear CCD array 52 to calibrate the r ader, ~ig. 24. Naster calibration document 150 includes a first vertical calibration line 152 corresponding to the left edge 126 of the fir~t column of 10 marking areas 95. Second vertical calibration line 154 corresponds to left edge 127 of the second line of marking areas 97. Third vertical calibration line 156 corresponds to edge 129 of character area 92. First horizontal calibration l:ine lS8 corresponds to edge 128 of a first row 15 of marking areas 131. Second horizontal calibration line 160 corresponds to edge 101 of a second row of marking areas 133.
As~aster calibration document 150 is transported across window 42 of reader 20, the master calibra~ion 20 document 150 is read by linear C~D sen~or 52. While the master calibration document ~50 is forced against side rail 29 by bias drive rallers 28, calibration document 150 is scanned beyond both edges of the document~ Thi~ is illustrated with regard to Fig. 5 where pixel areas scanned 25 extend beyond the æide edges of transaction document 90.
As generally described in Fig. 25, after the i~age data representative of the calibration document 150 is captured and stored, it is prob~d to locate the edges and lines of the calibratio~ document 150 in image memory 30 62. The edge of the calibration document against side rail 29 and is found by probing at least one row of single bit signals from one side of the image ~emory ~2 to locate at least one ~ingle bit ~ignal representative of a light pixel area. In ~ similar manner, the edge opposite 108 the rail 35 edge 106 i~ located in memory for the master calibration document 150 by probing the image memory from the other side. From either of these ed~es, the first, second and SUE3STlTUT~ SHEET

"; ~' W093/05480 . . . PCT/US92/072~ : :; :
lg 2116677 ~ ~
third vertical calibration lines 152, 154, 156 are located in memory and establish the location of marking areas 94 and character area 92 to provide ~or correction of geometric distortions created by the reader 20 as will be 5 discussed later.
Likewise, the leading edge of master calibration document 150 is located in image memory 62 by probing at least one column of memory to locate a first row of single bit signals representative of the leading edge 104. The 10 first and second horizontal calibration line, 158 and 160, are located accordingly ~o provide for correction of motion variability compensation as will be discussed later.
In addition, the horizontal and vertical resolution of the reader, pixel areas per inch, is 15 determined. ~his provides additional information for allowing scaling correction of mea~urements made wh~n probing image memory. The data from the reading of the master calibration document 150 is storsd in a non-volatile portion of memory 66, thus calibration need not be 20 performed again until the reader characterictics change.
Prior to both calibration of the reader ~0 and thresholding o~ transform memory, microscopic measurement o~ master calibration document lS0 and other portion~ of ~he transaction document 90 as needed are made, Fig. 24.
25 Microscopic ~easurement of the ~aster calibration docu~ent 150 is used to establish predetermined distances from the edges of the ~aster cali~ration document 150 to the marking areas 94 and character area 92. The microscopic measurements are stored in non-volatile portions of ~emory 30 66. Thus, when interrogating or probing image ~emory 62 a~ter rows of pixel.areas 102 have been scanned by tr~nsaction document reader 20, ~arking areas 94 and character area 92 can be located in ~he image memory 62 per the stored distances after the edges have been found~
After the transaction document reader 20 includPs stored microscopic measurements, is calibrated via master calibration document 150 and transform memory 72 is mapped, ' SIJEIST17~1~E SI~

W0~3/05480 ~I 1 6 6 7 7 PCT/US92~0726 image memory 62 can be probed or interrcgated to identify whether handwritten transaction marks are existent in the marking area~ 94 and interrogated to r~ad characters in character area 92 dyn~mically a~ tran~action document 90 i~
5 being scanned. The characters are of a predetermined ~et of characters, such as o-gO To determine whether a transaction mark 96 exists in marking areas 94 on tran~action document gO, CCD array timing control 64 notifies processing and control circuitry 58, in 10 conjunction with the system clock, when sufficient rows of pixel areas 102 representative o~ image data on document 90 have been ccanned, transformed, and stored in image memory 62 for interrogation of image memory 62 to beyin. When, for example, single bit signals representative of thirty 15 rows of pixel areas have been stored in image memory 62, the first row of marking areas 131 can be read. The number of rows needecl for the start of interrogation can be easily reprogra~med.
Fir~st, ~he captured rows and columns in image 20 memory 62 are probed to locate the edges o~ the document 90, Fig. 24. The rows of bits 134 in image me2ory 130 are interrogated ~y row probe 142 which can probe three rows of memory portion 130 simultaneously. Image memory 130, Fig.
8, represents a portion of image memory 62 corresponding to 25 the enlarged portion of transaction document in Fig. 7.
Rows of pixel areas can be interrogated from both sides of the entire image memory 6~. When ~wo or more consecutive light pixels are located by probing the image memory 62 from the sides of the image memory, both edges, 30 rail ide edge 106 and opposite edge 108, of the transaction document 90 ar~ located. The width of the document is directly translatable from the number of pixels between the edges 106, 108. The type of transaction document being read is identified from the document's 35 dimensions as each type of transaction documPnt is of certain dimensions.
Likewise, ~he columns of bits 132 within image Sl 1~S ~ iTUTE ~EET

: : ' W093/0~80 PCT/US92/07260 :
21 21~ ~677 . . .~ .. .
memory portion 130 are interrogated by col~mn probe 140.
When two or ~ore consecutive light pixel~ are loca~ed by probing the single bit ~ignals in the first several rows of image memory, ~he leading edge 104 of the transaction 5 document 90 is located. Once the edges 106, 108 and leading edge 104 of the document 90 have been located, the portion of image memory 130 representative of the marking area 94 of transaction document 90, represented in Fig. 8 as the phantom image memory marking area 140, can be 10 located by use of the microscopic measured predetermined distances ~or that type of transaction document identified stored in non-volatile portions of memory 66.
The marking areas 94 are shown as phantom lines as they are nok visible in image memory. They are printed 15 in a color which when it reflects light is above the threshold val.ue and is represented as a one bit signal or a light pixel area. Only the dark pixel areas are represented by the "X" in image memory 130, For example, red marking boxes will be invisible when illuminated by red 20 light sources.
The reading of ~he mark area is performed as generally de~cribed in Fig. 26. When the portion of image memory representative of the marking area has been located, phantom image marking area 144, that portion of i~age 25 memory is interrogated by row probe 142 to determine if a mark 96 exists therein. First, a number of r~ws (R~ of single bit ~ignals in image ~2mory 130 representative of the marking area 94 are probed by row probe 142. For example, ~ive rows o~ the twenty rows may be interrogated.
30 A predetermined number of rows, (P) three rows in the preferred embodiment, wherein hits are located (a hit being the location of at least two consecutive ~ingle bit signal~
representative of dark pixels) is established. As the five rows of single bit signals are being probed wi~hin the 35 phantom image memory ~arking area 144, the rows wherein hitæ occur (count) are counted and a filtering procedure is applied. If the number of rows where hits occur (count) - -SUE~STITUTE SHEE~

W093/0~80 PCT/US92/07260 21 1 6~77 22 is greater ~han or equal to ~he predetermined number (P), then mark is identified as being located in ~hi~ marking area. If no hits are located within the row~ of ~ingle b~t signals, than a mark doe~ not exi~t in the marking area.
s If, however, one or more, but les than the predetermined number (P) of rows of single bit signals within phantom marking area 144 are found to contain hits, thi~ marking area is considered a marginal marking area.
The marginal marking area i8 further filtered to 10 determine if a mark exi~ts therein. One method u~ed to filter the marginal area is by counting the number of single bit siynals within the phantom image memory marking area 144 of image m~mory 130 which are representative of dark pixels. I~ the number of dark pixels is greater than 15 a predetermined number established by a predetermined percentage of all pixel areas, for example 35 in a 20 by 20 pixel marking area, ~hen a mark exist~ therein.
In another embodiment of the invention, the filtering of t~e marginal maxking area is accomplished by 20 applying a neural network to the portion of image memory representative of the marking area to determine whether a mark QXi8tS therein. The neural network is trained to recognize such a mark in the manner neural networks are normally trained and as known in the art.
If all the row~ of single bit ~ignals probed locate a hit, then it is possible ~hat the user who made the handwrltten mark att~mpted to cratch out the mark and the ~ntire marking area may be represented in memory by dark pixel~. To filter this type of marking area, one 30 method used is to count all the dark pixels which appear in the~ marking area 144 and compare the count to a predetermined percentage o~ oYerall single bit signals in the-marking area 144, for example, 92~. If the count is larger than ~hak p~rcentage~ then ~he user scratched out 35 the handwritten mark.
In another embodiment of khe invention, ~he filtering of a possible scratched out mark is accomplished su~lt~)~ ~1 i~

W093/0~80 23 2 1 ~ ~ 6 PCT/US92/~7260 by applying a neural network to the portion of imag~ memory repxecentative of the marking area to determine if a mark ~:
i~ scratched out. ~he neural network is trained to recognize uch a scratched out mark in the manner neural 5 networks are normally trained and as i~ known in the art.
Further, although the above description probes rows o~ 5ingle bit signals, the probing within the image memory marking area can also be done on columns within the marking area. Even diagonal lines of single bit signals -10 within the marking area can be probed to read the mark. For example, five diagonal lines of ~ingle bit signals can be probed and the count o~ hits in these lines can be filtered. Thus, any line of pixel areas across the marking area can be probed.
l~ is readily k~own in the art that any number of pixel areas c:an be set to designate a hlt, any number o~
rows may be probed to determined if a mar~ exists, a row or ~ ~`
column probe can probe any number of rows or columns simultaneously, and any number. of consecutive pixel areas 20 whether dark or light can be chos~n when attempting to conclude that an area or edge of the document has been ~ ~
~ound. For example, the number of rows of single bit ~- :
signals in phantom Lmage memory marking area 144 interrogated to determine whether a mark exists could 25 easily be ten or three instead of five. ~ ~-With reference to Fig. 27 the optical reading of characters is described. The character area 92 of transaction do~ument 90 i5 located by means of the edges of the document and predetermined di~tan~es established by the 30 stored microscopic measurements in the same manner as the - :
marking areas 94 were located, as discussed above. Once a portion of the character area 92 encompassing ~he character "4", enlarged in Fig. 10, is stored in image me~ory 62, ~he ~ingle bit signals r~pre~antative of pixel areas are 35 interrogated therein to recognize the character.
Fig. 11 illustrates the probing of image memory 62 to define a specific area 110 wherein a single character : :
:
SUBS~lTlJ ~ F SHEET

W093/0~80 PCT/US92/072~
21:~6677 24 166 is located. Image memory portion 108 i8 probed frsm the leading edge of the tran~action document so by column probe 168. If the column probe does not locate a ~ir~t hit (two conce~utive fiingle bit signals representative of S dark pixel), then columns on both sides of those previously probed are interrogated until such a column with a first hit is located. The single bit signals located by ~he fir~t hit are utilized as an initialization point.
From the initialization point, columns adjacent 10 these initial single bit signals are interrogated to find hits adjacent to the initialization point. As the probes are moved, column to column, hits being continuously located on both sides of the initialization point, the sides of a specific area 110 are defined at the point where 15 no ~urther hits are located. In a like manner row probe 170 is utilized to probe the rows adjacent the ~ingle bit signals at the initialization point to establish an upper and lower se~ment of specific area 110. Thus, a specific area 110 encompa~s the entire single character 166. The 20 ~pecific area 110 located within image memory portion 108 is then compared to a predetermined area to acce~ ~hether the size of the 8peci~ic area is large enough to encompass a single character. If not large enough, the reader tries to locate a new specific area.
Aftar the ~pecific area 110 is located, it is normalized in a manner as is readily known in the art. Fig.
12 is the ~pecific area 110 wherein the character "4" has been norm~liz~d. Ea6h ~'X" 178, Fig. 1~, repre~ents stor~d signals in ~emory 66 of ~hether a certain area of image 30 memory 108 contains a predetermined number of single bit signals r~presentative of dark pixels.
Once the specific area 110 has been norm lized and stored in memory 66, a neural network is applied to the normalized area to determine the identity o~ the single 35 Gharacter. A limited number of characters, for example 0 through 9, are utilized in the invention to facilitate simple training of the neural network. In the training, ~ t~ur~ s ~ ~ ' .

W093/0~0 P~T/U~92/07~60 211~677 for example, the neural network i8 expo~ed to approximately 1,000-3,000 normalized pattern~ repre entative of a nu~eral ~'4", floating point decimal is eliminated, and back propagation i8 u~ed to adjust the network if it does not s ~ecognize th~ proper character. Su~h training of neural networks is readily known in the art. When applying th~
neural network to the normalized first character 176, Fig~
12, only a two layer neural network is utilized.
After a first specific area ~10 is located, a 10 second specific area, which would encompass the numeral "1", is easily located by dropping a column probe from the specific area 110 down the column~ until a hit is located in image ~e~ory which would be representative of the next character. Thus, the character area 92 need not be located 15 again, and a ~;pecific row of columns is probed to find the initialization point of the next character "1".
Whi:Le it is being determined whe~her a ~ark exists in the first row of marking areas 131, further pixel areas are being sensed, transformed into single bit signals 20 and stored in image memory 62. Thus, transac~ion document reader 20 is clyna~ically capturing the data on tran~action document 90 a~; it is transported through khe reader 200 By performing the identification of marks 96 in the ~irst row of marking areas 131, while ~imultaneously transforming and 25 storing further single bit signals, the ~peed of reading the marking areas of transaction document gO is greatly increased.
~ ikewi e, a~ the single bit signal~
r~pre~entative of the character "4" are being interrogat~d 3Q and ~he character recognized, the pixel areas repr~sentative of the character "1" are baing ~ensed~
transformed into single bit ~ignal~, and stored in im~ge m~mory 62. Once again, this dyna~ic capture and optical character recognition of these characters increases the . 35 speed upon which several characters on a tran~action document can be read.
Fig. 6 shows a receipt 120 which is read by SUBSTl`rUTE StlEET

W093/05~0 PCT/US92~07260 transaction document reader 20. The dimensions of the receipt 120 can be determined by probing the row~ and columns of single bi~ signals r~pre~entative of the pixel areas in the same manner a~ probed with regard to 5 transaction document 90. Th~ dimensions o~ receipt 120 correspond to a specific receipt docu~ent which the transaction document reader 20 will identify from predetermined stored in~ormation concerning its dimensions.
The reader is able to distinguish between the transaction 10 document 90 and the receipt 120 by determining the width or dimensions of the document.
Only a few rows of single bit signals need to be probed to determine the width and as such to deter~ine the type of document being read. Therefore, only a small 15 portion of the document is read prior to identifying the document to facilitate furthex reading of image data on the document. As shown in Fig. 24, when the receipt document is identified the orientation of the document is determined from the receipt dimensions. The characters ~0 representativle of the receipt number 124 are read in the same manner as predetermined characters 104 were read as discussed above with regard to transaction document 90.
Thu~, ik is recognized that the types of documents which can be read by transaction document reader 20 are numerous.
Figures 13-16 illustrate the use of the transaction document reader 20 with a tran~action docu~ent 180 having a certain number of strobes or timing marks 184 therson. Although the r~ading of tran~action documPnt lB0 is ~imilar to the r~ading of ~ransaction document 90, the 30 use of strobes or timing mark~ 184 provides a different marking area and character area location proce ~. ~he doc~ ent 180 of Fig. 13 includes marking areas 182, marks 186 therein, and a character area 183, all of which are similar to transaction doc~ment 90. ~he difference between 35 transaction document 90 and transaction document 180 i~ the inclusion of timing marks 184.
To calibrate the transaction document reader 20, SUBSTITUTE SHEET

W093/0~80 .i , PCT/US92/072~
27 21~ 6677 : :`
master document 200, Fig. 16, is transported ~hrough the reader 20, sensed, transformed, and ~tored to provide an image map of transaction document 18~. Master calibration document 200 includes calibration marks 202, a first 5 vertical master line 204 which correspond~ to the left edge of a first column of marking areas 177, and a second vertical master line 206 corresponding to the left edge of a second column of marking areas 173. Horizontal lines as used with transaction document 90 are unnecessary as the 10 top and bottom of a marking area can be located by counting the number of pixel rows from the leading edge of the document within the image memory, and vertical or motion variability compensation can be accomplished by means of the timing marks instead of horizontal ~alibration lines, 15 as shall be d:iscussed ~urther below.
Microscopic measurement data of the transaction document 1~0 a~re stored in memory ~o that the marking areac lB2 and character areas 183 can be located with respect to the timing marks 184 on the document. The microscopic 20 measurements are made between ~he timing marks 184 and the marking areas 182, and also with regard to the leading edge 203 and marking areas 182.
As discussed prQviously with regard to transaction document 90, the sides of transaction document 25 180 are located to determine the width of the document which corresponds to a specific type of transaction document. The image ~emory is furth~r interrogated to locate ~he leading edge 203 of the transaction docu~ent 180 by probin~ several columns of several rows of image memory 30 with column probe 192, Fig. 1~. Once the leading edge 203 of the document in image memory 62 is located, column probe 192 probes a predeter~ined number of columns wherein transformed timing mark 190 is represented by single bit signals. The probe l92 is co~monly a three column probe.
Once a hit is looated representative of a fir~t mark 185, columns adjacent ~hereto are probed to locate the left edge of the mark. After the left edge is located ~he S~;TtT~ E SHEET

w093/~5480 2 ~ ~ 6 6 7 7 PCT/USg2/0726~

left edges of the marking boxes can be found ~rom the predetermined microscopic measurements ~tored in memory.
The other ~ides of the phantom image memory marking area ~96 can be ascertained from stored measuxements from the S leading edge 203 of the document 180 and the dimensions of the marking area 196 itself. As previously described, now that the phantom image memory marking area 196 located within image memory portion 198 is found, further interrogation of the single bit signals within said phantom 10 image memory marking area 196 is performed to determine if a mark 182 exists therein.
Once a ~irst timing mark 185 is located in image memory and the left edge of the timing mark 185 i~ found, the center of the strobe can be found from known stored 15 dimensions of the timing marX. ~ ~econd timing mark 187 can be located in image memory by dropping a probe from the first timing mark center and down columns therefrom. ~hen the second ti:ming mark 187 is located, the marking areas in the row corresponding to the second timing mark can be 20 located withi.n image memory 62. Further, the characters within charaater area 183 are read in a manner ~imilar to the locating and reading of the character area of transaction document 90.
Documents with pre-defined row or column timing . 25 marks or ~trobes, may be processed with the process as described above with reference to document 180~ These documents ~ay also be processed without reference to the pre-defined tLming marks as described above with reference to transaction document 90. The pre-de~ined timing marks 30 are optional.
. In order to correctly identify the marks and character~ on the transaction documents, compensation for various distortions are provided by adjusting the probing of image memory 62, as generally shown in Fig. 26 and 27.
35 First, geometric diskortion as a result of the sensing device 34 scanning the rows of pixel areas of a transaction document is compensated. Some of the distortion is cau~ed .: '.:
St~.~S~

W093/0~80 29 ~ 2 1 1 ~ 6 7 7 pcT/uss2m by the len~es used therein1 Figure 18B illustrat~s ~he distortion. ~he actual marking ar~a 226 corresponds to ~he marking area as measured on a tran~action docu~ent.
The dashed i~age memory marking area 22~ ccrresponds to the 5 portion of image ~emory representative of the actual marking araa 226. As illustrated, ~he marking area in image memory 228 may be offset from the actual ~arking area 226 of the transaction document.
The master calibration document lS0, Fig. 9, is.
10 utilized to gather data to compensate for this geometric distortion cau~ed by the lenses of the sensing device 34.
The first, second and third vertical calibration lines 152, 154, 1~6 are utilized to properly locate the marking areas within the image memory. For example, if when the 15 transaction clocument reader 20 is calibrat~d, the first calibration line 152 is located 100 pixel areas from the edge of the I~aster docu~ent 150 as ~hown in image memory 62, then if the actual distance on the master document is 80 pixel area~s, a correction factor of 20 would be utili2ed 20 to properly locate the ~ingle bit signals repre~entative of the pixel areas within the marking area.
The actual bits within image memory 62 are not adjusted. Rather, the correction factor i6 utilized when interrogating the image memory 62 storing data of 25 transaction document 90 such that, as the column probe 140, Fig. 8, are interrogating the marking area, the proper single bit 8ignal5 within the image memory 62 are probed.
By adjusting the probe rather than restructuring and recalculating the image m2mory, compensation time is 30 reduced. For example, if a column probe was to probe the single bit signal representative of the 80th pixel area from the edge 108 of the transaction document 90, the single bit ~ignal repre~entative of the 100th pixel area would be probed in tead. The same geometric distortion 35 correction applies with regard to the reading of transaction document 180 as applies to transaction document ' 90.

SUBSTITUT~ ~HEET

W093/0~80 I PCT/US92/072~
2~6677 30 Distortion along the colu~ns of pixel areas is creat2d by the motion variability of the transaction document 90 as it is transported across window 42 o~ ~he: ~ :
sensing device 34. This di tortion is shown by Fig. 18C. ~;
S The actual marking area ~24 may be slightly o~fse~ from ~he marking area in image memory as represented by the da~hed line 222.
::
When reading the transaction document 90, the first and second horizontal calibration lines 158 and 160, 10 Fig. 9, provide for motion variability compencation in the vertical direction in a manner similar to geometric distortion compen~ation.
There are a predetermined number of pixel area~
between the leading edge of the actual ma~ter document and 15 the first horizontal calibration line 158. When ~he master document 150 i.s transported and ficanned by the transaction document reader 20, the first horizontal calibration line 158 is determined to be of~set in image memory 62 a certain number of pixl31s from its actual pixel distance from the 20 leading edge. As æuch, the difference between the actual pixel areas a~nd the pixels in image memory provide~ a correction factor ~or adjustiny the probes as the image memory 62 i~ interrogated, such that ~he row probe 142 probes the correct cingle bit signalæ in the columns of 25 single bit signalQ representative of the pixel areas within marking area 94. For example, if there i~ a correction factcr of 20 and the row probe was to probe ~he 80~h pixel in a column ~rom the leading edge of the docu~ent, the row probe would probe the 100th pixel instead. 5uch 30 co~pensation is al~o provided by the second horizontal calibration line 160.
When reading a transaction document having a number o f timing marks 184, Fig. 13, ~he ~pacing between the timing marks 184 is utilized to compensate for motion 35 variability di~tortion while transaction do~ument 180 is transported, and rows of pixel areas are sensed, transformed and stored in image memory 62. For example, ~ -- .
SlJB~ ~ S~

W093/0~80 PCT/US92/~72~0 31 2~ g6~7 the actual distance between the timing marks i~ known and ~tored a~ a certain dista~ce. When the actual tran~action document 180 is read by the transaction document reader 20 the space between the timing marks may b~ a different 5 n~mber of pixel areas than that on the actual documQnt.
As such, the difference between the number o~ pixel area~
in the image memory and that as determined by me~surement of the actual document defin~s a correction factor. The correction factor is then utilized to adju~t ~he probing of 10 the rows of singl~ bit signals in the image memory to properly probe the correct rows representative of the marking area 182.
Although the bias drive roller 28 forces the transaction document 90 against the side rail 2~ of the 15 document reader such that misalignment is virtually eliminated, if the document 90 is not transported through the reader 20 such that the leading edge 104 is parallel to the linear array ~2nsor 52, skew distortion will occur.
Figure 17 and 18A illustrate this distortion. For example, 20 actual marking area 22g will be skewed in the image memory 62 relative to i~s unskewed position, the dashPd line area 218. The ~k~ew of the document 90 as it is transported through the reader 20 can be determined by reading the N
and N distances as shown in Fig. 17. An M/N ratio is used 25 to adjust the probing of the pixel areas so that interrogation of correct single bit ~ignals in imag~ m~mory 62 provided. This adjustment is ~ade for both the column probe 140 and the row probe 142 as ekew distorts both rows and colu~ns of pixel areas.
In addition, the amount of skew can be deter~ined by the deviation of the rail edge of the document in image ~emory from the location of calibrated rail edge o~ the ~aster document as stored in ~emory 66. This deviation, or triangle formed by a skewed document, as is shown in 35 Fig. 17 along line 2~0 can also be used to calculate skew correction.
If a transaction document includes marks 184, SUBSTIl U~F SHEÇ~T

W093~05480 2 1 ~ ~ ~ 7 7 PCT/US~/072~0 such as transaction document 180, then the slope of the ~ :
marks 184 in image memory can be used to determine ~he adjustment ratio.
It ls emphasized that the addre~ses of the sin~le -~
S bit signals are not changed within the image memory 62 to c~mpensate for any of ~he previously mentioned distortions.
Rather, the probing of the single bit signals in image memory 62 under the control of software 68 is adjusted ~y the correction factors as discussed above. By adjusting 10 the probing rather than the actual memory, cuch compénsation is accomplished on a pixel by pixel basis as needed, rather than awaiting adjustment of an entire image memory, thus reducing the time r~quired for such compensation.
In alnother embodiment of the invention, the light sources 38 of sensing device 34, Fig. 2 and Fig. 4, are red and green light sources, preferably red ~nd green LED's.
With u~e of the red and green LED's, red and green printed data becomes visible in image memory 6~, thus allowing red 20 and green features on transaction document 90 to be read.
The red and green LED's are operatively positioned to uniformly illuminate the entire width o~ a transaction document transported across window 42 with each of the colors, red and green. The LED's are switched 25 approximately every ~ew hundred microsecond, enabling the selection of the color of illumination in ~he interval between succe~ive scans and sensing of rows of pixel areas 102 of the doc~ment as it is being transported acro~ the window 42~ ~he red and green LED's alternately illu~inate 30 the rows o~ pixel areas 102 under the control of CCD array timing control 64.
The transfsrm memory 72 ie mapped with red and green threshold mapping. Ones and zeros are ~apped in a similar manner as when only dark and light pixels were 35 considered. The tran~form memory is a 32R ~emory organized 256 x 128. With red/green mapping, the upper nibble is -~
used for green threshold mapping and the lower nibble used :.~
~ ~ .

SVBSrlrVTE SHEET
.: -~ ~ .
~.

W093/0~80 2 ~ 1 ~ 6 7 7 PCT/US92/072~ ~

for red mapping. The reading o~ a plain reflectiYe document once again mapc ~he transform memory with ones and zeros. By averaging the gray ~cale values and applying a pradetermined threshold percentage, a threchold value i~
5 determined and the addresses are loaded accordingly.
When a transaction document prin$ed with red and green is read, both the red a~d green light emitted from the respective LED's will re~lect back from white or light printing to the linear CCD ensor 52 at a level above the 10 respective threshold values for the green and red illumination such that output signals representative of light pixel areas are provided. Green light is re~lected back to the CCD sensor 52 ~rom green printing of the correct wavelength at a value above the green threshold 15 values, and output signals representative of light pixels are provided. The same green printing does not reflect light above t]he red threshold value when illuminated with the red LED's light. Therefore, an output signal representative of a dark pixel is provided. Red light is 20 reflected baclc to the CCD image ~en~or 52 from red printing of the correct wavelength at a value above the red threshold values. ~he same red printing does not reflect light above the gr~en threshold value when illuminated by the green LED's light. Therefore, output signals 25 representative of dark pixels are provided when red printing is illuminated by green LED's. Black printing or dark printing not containing ~ither green or red is not reflected above the respective threshold value back to ~he CCD linear image sensor 52, and thus output signals 30 representative of dark or black pixel areas are provided whether the black printing is illuminated by green or red LED's~
The LED lighting is synchronized such that rows of pixels 102 which are even ~umbered in the imagC memory 35 62 are always illuminated with red and, conversely, rows of pixels 102 which are odd numbered in image memory are always illuminated with green. Thus~ as shown in Fig. 19, SU~S~ E St~EE

W093/05480 2 1 1 6 ~ 7 ~ PCT/US92/07260 a green field 232 which makes red printing ~isible and a red field 234 making green printing ~isible in image ~emory 62 is crea~ed. As long as the lighting is ~ynchronized, proper image processing can be performed. Therefore, as 5 one probes through the memory array, row by row, which color LED was the cause of any reflection that was above the threshold values is determined. Conversely, it i~
determined which color is absent, including white, in the absence of any ligh~ abov~ the threshold val~e for that lo color LED.
In addition ~o synchronization of the input of the image memory 62 with respect to the color LED
illuminating the document, it is also necessary to maintain synchronization of the transform memory 72 so that the 15 addresses mapped according to the threshold values for a column in the row of pixel areas 102 are those mapped for the correspon~ding color LE~.
When the transform memory is mapped, the capture of image data from the document can be accomplished. The 20 7-bit intensity output signal ~rom A/D converter 70 for a pixel area in a row of scanned pixel areas, a bit position signal from transform control 74 representative of the column number of the pixel area, and a second bit signal from trans~orm control 74 representative of the nibble 25 which was used for mapping the particular light source color, look up for the pixel area whether the pixel area i~
to be represent~d by a "1" or "0". As discussed with the ~ingle color liyht source, the image me~ory tores the~e single bit si~nals which are generated on a pixel by pixel 30 basi~
A black ~arking box 230 illuminated alternately with a red LED and a green LED i shown in Fig. 19. The red illuminated rows of pixel areas create a red field 234 and the green illu~inated rows of pixel areas create a 35 green field 232. ~ny red printing is captured in the green field 232, and green printed images or features are captured in the red field 234.

SUE3STI~UTE SHEET
- - .

W093/0~ P~TtUS92/07~
2~677 ~ ra~itionally, by use of filter~ and other techniques, document ~ields ha~ing different colored printing have been illuminated by topologically isolating color light sources, or illuminated by a broad spectrum 5 light sourc~ and topologically filtering the reflected light. For examplP, one quarter of the document printed in green would be illuminated by red filtered light, and three quarters of the document printed in red would be illuminated by green ~iltered light. Such topological 10 isolation of color light sources upon a particular docuoent leads to inflexibility in design and mechanical and illumination problems at the transition between the red printed portion of the document and a green printed portion of the document. By alternately illuminating the rows of 15 pixel areas of the document, the problems of inflexible design and trilnsition boundary are eliminated.
By providing a red and green image field ~hrough alternate ill~mina~ion of the document, red and green printing can be placed anywhere on the documentO Thi~
20 flexibility l~ads to a document topological layout ~hat can be aesthetically pleasing, in that the color~ no longer need to be topologically separated in a spacial layoutO
In addition, numerous choices can be made as to what ~hall be visible and invisible in the image memory 62. For 25 example7 the marking area could be printed in red and thus become visible by illumination with green light, while ~trobes could be printed in green and thu8 become visible in the- red field 234.
Fig~r. 20A illuctratec a black ~arking-area after 30 illumination by alternatlng green and red LEDs, transformed, and stored in i~age memory 62. The black features are seen in both the red and green image fields 234, 232 as neither the red or green light will reflect above the red or green threshold values for the red and 35 green fields 234,232. Fig. 20B illustrates in red field 234 the black marking area. Fig. 20C hows the black . marking area in green field 232. Black features do not ~13BST8TU~ ~

w093/0~80 2 1 1 6 6 7,7 PCT/US92/072~0 reflect light above the threshold values of the green or rQd LED's and are therefore, visible in both fields.
Tha fl~xibility of utilizing various printed color and ~he alternating green LED and red ~ED ~-5 illumination of the document, with those printed color~
thereon, is demons~rated in Fi~s. 21 and 22. In Fig. 21, the marking areas 244 are printed in green and the timing marks 246 are printed in red. As the green marking area and red timing mark are sensed, transfor~ed and stored in 10 a portion of image memory 62, Fig. 22, tha single bit signals of tha transformed green marking area 248 can be probed in the red field such that the marking area 244 can be located in image memory. This facilitates location of the marking area in image me~ory. Handwritten marks which 15 are made therein in black would appear in both the red and green fields and single bit signals representative of ~he mar~ing area are probed to determine if a handwritten mark ~:
exists therein. In addition, the green field could be probed to ~ind the transformed red timing mark 252.
To identify whether a handwritten mark exist~ in a marking area, either the red or green field can be probed, rather than the whole box as discussed previou~ly, when describing the reading of transaction document 90.
If a marginal marking area is located, only the pixel areas 25 of either of the fields ne~ds to be counted to determine if a mark exists. This reduces the amount of time required ~ ~:
to recognize handwritten marks in the marginal areas. ;;~
In another alternative embodiment, ~he - ~ .
recognition of multiple color printed data on ~he :~
30 transaction document is accomplished by having a full spectrum white light .ource illuminating ~ultiple colors. :: -The refl~cted light is then sensed by a color CCD sensor, ~;~ ;;:
Fig. 28. This eliminates the need for switching the LED~s -~
as di~cus~ed above and provides even greater ~lexibility of 35 document de~ign. The color CCD provides three outputs for ~ach pixel area, one for each of the colors, blua, green and red. -':

SUBST~3:~E S~tEET ;

W093/0~80 2 ~ 1 ~ 6 7 7 PCT/US92/07260 Transform memory 72 is ~imilar to the transform memory utilized in the alternating red and green embodiment, except the trans~orm memory includes three portions instead of two, one for each o~ the three color , 5 red, green, and blue. Each portion o~ the transform memory 72 is mapped in a similar manner as done for the other embodi~ents with plain reflective document. When the reflective doc~ment is read by ~he document reader, an average of the digitized intensity output signals with a 10 sensitivity of 128, representative of each o~ the red, yreen and blue olltputs from CCD sensor 250, is generated for each color along the columns of the re~lective document. Predetermined percentage~ for each of the red, green and blule colors are applied, in a similar manner as 15 done with respect the previously described embodiments, to each of the red, green, and blue averaged digitized intensity.out]?ut signals to determine the threshold values for each of the blue, red and green colors. The corresponding portions of transform memory are then mapped 20 with l's at addresses above or equal to the threshold value for each of the red, green and blue outputs and 0~8 at addre~#es below the threshold value for each of the red, green and blue outputs.
When a multi-colored document is then read, the 25 color CCD sensor 2S0 provide~ three output signals, red, green and blue, representative o~ the reflected light from each pixel of the multi-colored document. The output signals are applied to A/D converters 252, 254, and 256, under control of transform control 74 via ti~ing 8ignal8 30 from CCD array timing control 64. The A/D con~ertors provide digitized intensity output Rignals, 7 bits, to address the transform memory mapped as a function of the threshold values for the three different colors-. The transform ~emory is also addressed by ~ignals, ten bits, 35 from transform aontrol 64 representative of ~he column o~
each pixel area, and signals, two bits, from ~he transform control 64 repre~entative of the three different colors.

~;UE3S~ITUTE SHEET

~093/05480 , i~ PCT/US92/07260 21~6677 38 A cingle bit ~ignal for each o~ the three colors is addreæsed for each pixeliarea and stored in imaga memory 6~. For example, if the digitized output ~ignals for a pixel area are above the threshold value for ~he blue color 5 but below the threshold values for the green and red colors, a "1" is selected for the blue output and ~O"s for the green and red colors, and the~ stored in image memory.
The single bit signals from the transform memory 72 are stored in separable image fi~lds, one for each color, or 10 three fields. Each of the color image fields are stored in the same memory array. These image ~ields are then intQrrogated in a similar manner as previously described herein with regard to the alternating light source embodiment.
The transaction document reader 20 can be utilized to facilitate transactions involving transaction document 90. Tran~action document 90 includeæ marking areas 94 to record mark~ to complete a tran~action. The transaction docu~ent 90 also includes ia signature/address 20 block 98. Usuially, in a normal transaction the signature and address block 98 of the transaction document 90 i~
required to ble completed. This information is typically necessary to associate a user with the transaction document. For example, the name is associated with a 25 winning lottery ticket. So that completing such address and name information iæ not necessary with the completion o each transaction document 90, a method for performing a transaction without æuch cignature may be acco~pli~hed with the use o~ reader 20. A facilitated transaction utili~ing 30 a lo~tery player shall be described in further detail below.
The lottery player is provided with a user identification card 254, Fig. 23A. The identification card 254 includes the player~ identification number 261.
35 The lottery player ~1 o has a player personalized document, such as a bank card with an account number thereon 258, Fig. 23C, or an address/æignature card 260, Fig. 23D. The : ~ , .

W093/0~80 .P~T/US92/07260 39 2P1~77 player personalized document ~ould be any document which when asso~iated with the player would identify the player.
A player, prior ~o making a first transaction, has identificat~on card 254 optically read by document 5 transaction reader 20. The personalized document, ~58 or 260, also is digitally imaged by the transaction document reader 20. The identification number 261 on identification card 254 is stored with the as~ociated digitally imaged player per~onalized document, 258 or 260. Therefore, in 10 subsequent transactions, the player per~onalized document, 258 or 260, is readily accessed by merely optically reading the identification number 261 with which the personalized document is associated. :
Thus, after such information is stored, lS transactions are completed by transporting a lottery document thr~Dugh the document reader, which reads the handwritten marks thereon and aleo tra~ports the user identificatia,n card therethrough so that the identification number 261 i~; optically recognized. The identification 20 number 261 is then associated with the personalized document information. As such, the player need not write his signature on the lottery document in a signature block like block 98, as the identification number 261 per~orms the function of associating the lottery document with the 25 player via the stored personalized document information.
This association function is also used to associate a receipt, Fig. 6, and the receipt nu~ber 124, to a personalized bank account nu~ber fr~m the bank account number card which was previously optically recognized and 30 previously stored in conjunction with identification number 261. Thus, if a lottery player had a winning receipt, he need only have the identification number on his identification card read and associated with the reading of the receipt ~umber in order to identify a bank account for 35 proceeds to be deposited within.
Further, reader 20 is provided with a card swipe 16 to read a magnetic strip 256 on the reverse side of SlJBST~TV~E SHET

W093/0~80 2 1 1 6 6 7 7 PCT/US92J072G0 identification card 254 with ~he ID card number 261 encoded ;~
thereon, Fig. 23B. Thus, either a magnetic reader or an optical reader could perform the reading function of the identification number and hasten the transaction.
s Although ~he present invention has been described above in a preferred ~orm, tho~e killed in the art will readily appreciate that various modifications can be mad~
without departing from the spirit and scope of the invention as bounded by the claims of the application 10 itself.
''.',-~; :', '., '~ ~

: . ~
~.'~:' ~..
. - . ~.-.,.

:

- :' ~

. - ~.
$~tE SI~E~f

Claims (69)

1. An apparatus for reading image data on a transaction document, said image data including a plurality of marking areas employed by a user to record marks, the apparatus comprising:
sensor means for sensing, one row at a time, a plurality of rows of image data as said document is transported across said sensor means, each row including a plurality of pixel areas, said pixel areas of said rows being aligned in a plurality of columns, said sensor means comprising means for generating output signals representative of said pixel areas;
transform means for receiving said output signals and generating bit signals representative of each pixel area;
image memory means for storing said bit signals, said image memory means being organized into rows and columns of bit signals corresponding to said plurality of rows and columns of said pixel areas of said image data;
means for locating at least one portion of said image memory means representative of at least one of said plurality of marking areas of said transaction document, said means for locating being operative after rows of bit signals representative of said at least one marking area are stored in said image memory means and while additional rows of bit signals representative of other image data are being sensed, transformed, and stored in said image memory;
means for identifying marks made in said at least one marking area while said additional rows of bit signals are being stored and while other marking areas are being located by probing said bit signals within said image memory means.

2. An apparatus according to claim 1, wherein said transform means comprises:
means for digitizing said output signals representative of said pixel areas and for providing gray scale digitized output signals;

transform memory means for generating and storing bit signals for individual columns as a function of a threshold value; and means for addressing said transform memory means to select a bit signal representative of whether the gray scale digitized output signals are representative of generally dark or light pixel areas.

3. An apparatus according to claim 2, wherein said means for addressing, comprises:
means for applying the gray scale digitized output signals representative of said pixel area and bit positioned signals representative of the columns of said pixel areas to said transform memory means, on a pixel by pixel basis; and means for selecting from said transform memory, in response to said means for applying, a bit signal for each of said gray scale digitized output signals representative of said pixel areas.

4. An apparatus according to claim 3, wherein said transform memory means comprises:
means for averaging a plurality of output signals representative of pixel areas in each column of pixel areas when a reflective card is transported across said sensor means and for generating an averaged output signal representative of said output signals;
means for applying a predetermined threshold percentage to the averaged output signal to determine a threshold value address; and means for mapping aid transform memory such that said transform memory contains a first bit signal at said threshold value address and at addresses higher than said threshold value address, and contains a second bit signal at addresses lower than said threshold value address.

5. An apparatus according to claim 1 wherein said means for locating, comprises:

means for identifying a type of transaction document by probing a predetermined plurality of rows of said image memory means in order to determine dimensions of said document, said type of document having certain dimensions; and means for locating said at least one portion of image memory means representative of aid at least one of said plurality of marking areas as a function of stored predetermined distances measured with respect to said dimensions of said type of transaction document identified.

6. An apparatus according to claim 1 wherein said means for locating, comprises:
means for identifying a type of transaction document by probing a predetermined plurality of rows of said image memory means to determine dimensions of said document, said type of document having predetermined dimensions;
means for probing at least one column of said image memory means to locate a first mark position, the image data of said document including a plurality of mark positions; and means for locating the at least one portion of image memory means representative of said at least one of said plurality of marking areas as a function of a predetermined distance from said first mark position to said at least one marking area dependent upon said type transaction document identified.

7. An apparatus according to claim 5, wherein said identifying means comprises:
means for finding a leading edge of said transaction document by probing said columns of said image memory as said rows of bit signals are stored; and means for finding side edges of said transaction document by probing said rows of said image memory such that, once side edges are found, the width of said document is realized.

8. An apparatus according to claim 7, wherein one of said side edges is found by calibration of said image memory, said one of said side edges corresponding to a side edge of said transaction document being forced against an alignment rail as said document is transported across said sensor means.

9. An apparatus according to claim 1 wherein said means for identifying marks, comprises:
means for probing a plurality of lines of bit signals representative of pixel areas extending across said marking area within said at least one portion of said image memory means in order to identify whether said lines contain at least one bit signal representative of a dark pixel area;
means for comparing the number of lines containing said at least one bit signal to at least one predetermined number of lines; and means for filtering the result of said means for comparing in order to identify whether a mark appears in said at least one marking area.

10. An apparatus according to claim 9 wherein said means for filtering comprises means for carrying out the following process:
if said means for probing identifies a number of lines less than a first predetermined number of lines but more than a second predetermined number of lines, counting the number of bit signals representative of dark pixel areas in said portion of image memory means representative of said at least one marking area and concluding that there is a mark in said at least one marking area if said number of bit signals counted is greater than a predetermined number of bit signals, otherwise concluding that no mark is present in said marking area;
if said means for probing identifies a number of lines greater than said first predetermined number of lines, concluding that there is a mark in said at least one marking area; and if said means for probing identifies a number of lines less than said second predetermined number of lines, concluding that there is no mark in said marking area.

11. An apparatus according to claim 9 wherein said means for filtering comprises means for carrying out the following process:
if said means for probing identifies a number of lines less than a first predetermined number of lines but more than a second predetermined number of lines, applying a neural network means to said portion of said image memory means representative of said at least one marking area, said neural network means for determining if a mark is placed in said marking area;
if said means for probing identifies a number of lines greater than said first predetermined number of lines, concluding that there is a mark in said at least one marking area; and if said means for probing identifies a number of lines less than said second predetermined number of lines, concluding that there is no mark in said marking area.

12. An apparatus according to claim 9 wherein said means for filtering comprises:
means for counting said bit signals representative of dark pixel areas in said at least one marking area whenever said means for probing identifies all said plurality of lines as having at least one bit signal representative of a dark pixel area; and means for identifying that a mark previously placed in the marking area has been scratched out by comparing aid number of bit signals to a certain percentage of overall bit signals representative of said at least one marking area.

13. An apparatus according to claim 1, further comprising means for providing timing control for reading image data as bit signals are being stored in said image memory means by notifying said means for locating and said means for identifying marks that said image memory means has a certain number of rows of bit signals stored therein so that reading of certain portions of said document can be initialized.

14. An apparatus according to claim 1, further comprising means for compensating for distortion across a row of pixel areas when the row is sensed by said sensor means so that probing of rows of said bit signals of said image memory representative of said at least one marking area is accurately performed when said marking area is located and marks therein are identified.

15. An apparatus according to claim 1, further comprising means for compensating for distortion along a column of pixel areas when a plurality of rows are sensed 80 that probing of columns of said bit signals of said image memory representative of said at least one marking area is accurately performed when said marking area is located and marks therein are identified.

16. An apparatus according to claim 1, further comprising means for compensating for an improper angle of said document as it is transported across said sensor means so that probing of said rows and columns of said bit signals of said image memory representative of said at least one marking area is accurately performed when said marking area is located and marks therein are identified.

17. An apparatus according to claim 6, further comprising means for controlling the speed at which said document is transported across said sensor means as a function of said plurality of mark positions of said document.

18. An apparatus for reading image data on a transaction document, said image data including pre-determined characters, comprising:
sensor means for sensing, one row at a time, a plurality of rows of image data as said transaction document is transported across said sensor means, each row including a plurality of pixel areas, said pixel areas of said rows being aligned in a plurality of columns, said sensor means comprising means for generating output signals representative of said pixel areas;
transform means for receiving said output signals and generating bit signals representative of each pixel area;
image memory means for storing said bit signals, said image memory means being organized into rows and columns of bit signals corresponding to said plurality of rows and columns of said pixel areas of said image data;
character locating means for locating a specific area which is within said image memory and which is representative of at least a single character, said character locating means being operative after said bit signals representative of said single character are stored in said image memory and while additional bit signals representative of additional image data are being sensed, transformed, and stored in said image memory means; and neural network means for recognizing said predetermined character within said specific area.

19. An apparatus according to claim 18, wherein said character locating means comprises:
first means for locating at least one portion of said image memory means representative of at least one character area of said document, wherein at least one predetermined character is positioned;
second mean for locating said specific area within said at least one portion of said image memory means representative of said at least one character area encompassing said at least one single character, said second means for locating being operative after said bit signals representative of said specific area are stored in said image memory and while additional bit signals representative of additional image data are being sensed, transformed, and stored in said image memory means.

20. An apparatus according to claim 18, wherein said transform means comprises:
means for digitizing said output signals representative of said pixel areas and for providing gray scale digitized output signals;
transform memory means for generating and storing bit signals for individual columns as a function of a threshold value; and means for addressing said transform memory means to select a bit signal representative of whether the gray scale digitized output signals are representative of generally dark or light pixel areas.

21. An apparatus according to claim 20, wherein said means for addressing, comprises:
means for applying the gray scale digitized output signals representative of said pixel areas and bit position signals representative of the columns of said pixel areas to said transform memory means, on a pixel by pixel basis; and means for selecting from said transform memory, in response to said means for applying, a bit signal for each of said gray scale digitized output signals representative of said pixel areas.

22. An apparatus according to claim 21, wherein said transform memory means comprises:
means for averaging a plurality of output signals representative of pixel areas in each column of pixel areas when a reflective card is transported across said sensor mean and for generating an averaged output signal representative of said output signals;
means for applying a predetermined threshold percentage to the averaged output signal to determine a threshold value address; and means for mapping said transform memory such that said transform memory contains a first bit signal at said threshold value address and at addresses higher than said threshold value address, and contains a second bit signal at addresses lower than said threshold value address.

23. An apparatus according to claim 19, wherein said first means for locating at least one portion of said image memory means representative of at least one character area, comprises:
means for identifying a type of transaction document by probing a predetermined plurality of rows of said image memory means to determine dimensions of said document, said type of document having predetermined dimensions, said at least one character area positioned at a predetermined location thereon; and third means for locating said at least one portion of aid image memory means representative of said at least one character area as a function of said dimensions and said predetermined location.

24. An apparatus according to claim 23, wherein said identifying means comprises:
means for finding a leading edge of said transaction document by probing said columns of said image memory as said rows of bit signals are stored;
means for finding side edges of said transaction document by probing said rows of said image memory such that, when side edges are found, the width of said document is realized.

25. An apparatus according to claim 24, wherein one of said side edges is found by calibration of said image memory, said one of said side edges corresponding to a side edge of said transaction document being forced against an alignment rail as said document is transported across said sensor means.

26. An apparatus according to claim 19, wherein said second means for locating said specific area within said at least one portion of said image memory means, comprises:
means for probing a first predetermined column of bit signals of said at least one portion of said image memory means representative of said character area;
means for probing additional columns on both sides of said first predetermined column whenever said first predetermined column does not include at least one bit signal representative of a dark pixel area, said additional columns being probed until a particular column with at least one bit signal representative of a dark pixel area is located, said at least one bit signal being part of said image memory means representative of said single character; and means for probing rows of bit signals of said image memory and for probing columns of bit signals of said image memory adjacent to and on both sides of said bit signal located in said particular column in order to locate said specific area encompassing said single character.

27. An apparatus according to claim 26 further comprising means for determining if said specific area located is of a predetermined size necessary for location of said single character.

28. An apparatus according to claim 18, wherein said neural network means comprises:
means for normalizing said specific area;
means for training aid neural network to recognize said predetermined characters; and means for applying said trained neural network to said nonrealized specific area to recognize said predetermined characters.

29. An apparatus according to claim 18, further comprising means for providing timing control for reading image data as bit signals are being stored in said image memory by notifying said character locating means that said image memory means has a certain number of rows of bit signals stored therein so that reading of certain portions of said document can be initialized.

30. An apparatus according to claim 18, further comprising means for compensating for distortion across a row of pixel areas when the row is sensed by said sensor means so that probing of rows of said bit signals of said image memory representative of said specific area is accurately performed when said specific area is located and characters therein are identified.

31. An apparatus according to claim 18, further comprising means for compensating for distortion along a column of pixel areas when a plurality of rows are sensed so that probing of columns of said bit signals of said image memory representative of said specific area is accurately performed when said specific area is located and characters therein are identified.

32. An apparatus according to claim 18, further comprising means for compensating for an improper angle of said document as it is transported across said sensor means so that probing of said rows and columns of said bit signals of said image memory representative of said specific area is accurately performed when said specific area is located and marks therein are identified.

33. An apparatus for determining a type of transaction document having image data thereon, comprising:
means for generating bit signals representative of rows and columns of dark or light pixel areas of said image data;
an image memory for storing said bit signals, said image memory being organized into rows and columns corresponding to said pixel areas of said image data of said document; and means for identifying said document as a function of said dimensions of said document after probing at least one row of said image memory to determine the documents dimensions.

34. An apparatus according to claim 33, identifying means comprises:

means for finding a leading edge of wherein said transaction document by probing said columns of said image memory as said rows of bit signals are stored;
means for finding side edges of said transaction document by probing said rows of said image memory such that, when side edges are found, the width of said document is realized.

35. An apparatus according to claim 34, wherein one of said side edges is found by calibration of said image memory, said one of said side edges corresponding to a side edge of said transaction document being forced against an alignment rail as said document is transported across said sensor means.

36. An apparatus for reading information on a transaction document having image data thereon, said image data including at least one marking area for a user to place a transaction mark, comprising:
means for generating bit signals representative of rows and columns of dark or light pixel areas of said image data;
an image memory for storing said bit signals, said image memory being organized in rows and columns corresponding to said pixel areas of said image data of said document;
means for locating at least one portion of said image memory representative of said at least one marking area as a function of particular reference characteristics of said document; and means for identifying whether said at least one marking area contains a mark or lacks a mark by probing a plurality of lines of said bit signals of said image memory representative of pixel areas extending across said marking area.

37. An apparatus according to claim 36, wherein said means for locating said at least one portion of said image memory representative of said at least one marking area, comprises:
means for identifying a type of transaction document by probing a predetermined plurality of rows of said image memory to determine dimensions of said document, said type of document having predetermined dimensions;
means for locating the at least one portion of image memory representative of aid at least one of said plurality of marking areas as a function of stored predetermined distances measured with respect to said dimensions of said type of transaction document identified.

38. An apparatus according to claim 36 wherein said means for locating aid portion of said image memory representative of said marking area comprises means for locating said at least one marking area by probing said bit signals of said image memory representative of a plurality of predetermined columns of said document and by locating at least one strobe mark, said marking area being a predetermined distance from said at least one strobe mark.

39. An apparatus according to claim 36 wherein said means for identifying comprises:
means for probing said plurality of lines of bit signals of said image memory representative of said pixel areas within said at least one marking area to identify whether said lines contain at least one bit signal representative of a dark pixel area;
means for comparing the number of lines containing said at least one bit signal to at least one predetermined number of lines; and means for filtering the result of said means for comparing to identify whether a mark appears in said at least one marking area.

40. An apparatus according to claim 39 wherein said means for filtering comprises means for carrying out the following process:
if said means for probing identifies a number of lines less than a first predetermined number of lines but more than a second predetermined number of lines, counting the number of bit signal representative of dark pixel area in said portion of image memory representative of said at least one marking area and concluding that there is a mark in said at least one marking area if said number of bit signals counted is greater than a predetermined number of bit signals, otherwise concluding that no mark is present in said marking area;
if said means for probing identifies a number of lines less than said second predetermined number of lines, concluding that there is a mark in said at least one marking area; and if said means for probing identifies a number of lines less than said second predetermined number of lines, concluding that there is no mark in said marking area.

41. An apparatus according to claim 39 wherein said means for filtering comprises means for carrying out the following process:
if said means for probing identifies a number of lines less than a first predetermined number of lines but more than a second predetermined number of lines, applying a neural network means to said portion of said image memory representative of said at least one marking area, said neural network comprising means for determining if a mark is placed in said marking area;
if said means for probing identifies a number of lines greater than said first predetermined number of lines, concluding that there is a mark in said at least one marking area; and if said means for probing identifies a number of lines less than said second predetermined number of lines, concluding that there is no mark in said marking area.

42. An apparatus according to claim 39 wherein said means for filtering comprises:
means for counting said bit signals representative of dark pixel areas in said at least one marking area whenever said means for probing identifies all said plurality of lines as having at least one bit signal representative of a dark pixel area; and means for identifying that a mark previously placed in the marking area has been scratched out by comparing said number of bit signals representative of dark pixel areas to a certain percentage of overall bit signals representative of said at least one marking area.

43. An apparatus for reading image data on a transaction document, said image data including predetermined characters located in at least one character area of said document, comprising:
means for generating bit signals representative of rows and columns of dark or light pixel areas of said image data;
an image memory for storing said bit signals, said image memory being organized into rows and columns of bit signals corresponding to said rows and columns of pixel areas of said image data of said document;
means for locating a specific area encompassing a single character within said character area by probing said rows and columns of said bit signals of said image memory representative of said character area;
and neural network means for recognizing said single character within said specific area.

44. An apparatus according to claim 43, further comprising means for locating said at least one portion of said image memory representative of said at least one character area, said means for locating comprising means for determining the width of said document by probing a predetermined plurality of rows of said image memory, said width corresponding to a particular type of document having said at least one character area at a predetermined location on said particular type of document, said at least one portion of said image memory being representative of said at least one character area located as a function of said width and said predetermined location.

45. An apparatus according to claim 43, wherein said means for locating said specific area, comprises:
means for probing a first predetermined column of bit signals of said at least one portion of said image memory representative of said character area;
means for probing additional columns on both sides of said first predetermined column whenever said first column does not include at least one bit signal representative of a dark pixel area, said additional columns being probed until a particular column with said at least one bit signal representative of a dark pixel area is located, said at least one bit signal being part of said image memory representative of said single character; and means for probing rows of bit signals of said image memory and for probing columns of bit signals of said image memory adjacent to and on both sides of said bit signal located in said particular column in order to locate said specific area encompassing said single character.

46. An apparatus according to claim 43 further comprising means for determining if said specific area located is of a predetermined size necessary for location of said single character.

47. An apparatus according to claim 43, wherein said neural network means comprises:
means for normalizing said specific area;
means for training said neural network to recognize said predetermined characters; and means for applying said trained neural network said normalized specific area to recognize said predetermined characters.

48. An apparatus for reading information on a transaction document having image data thereon, said image data including at least one read area, comprising:
a linear array sensor for sensing a plurality of rows of image data of said document, each row including a plurality of dark or light pixel areas, said pixel areas of said rows being aligned in a plurality of columns, means for generating bit signals representative of the dark or light pixel areas of said image data;
an image memory for storing said bit signals, said image memory being organized into rows and columns of bit signals corresponding to said rows and columns of pixel areas of said image data of said document;
means for calibrating said apparatus by transporting across said sensor a master document having lines at predetermined positions corresponding to said transaction document read area, and by comparing said predetermined position of aid master document lines measured with respect to document reference characteristics prior to transporting said master document across said sensor to the lines as represented by said bit signals in said image memory after said master document is sensed and said bit signals are generated and stored; and means responsive to the comparison of said calibrating means for compensating for distortion across the row of pixels sensed by said sensing means when said read area is located and read by probing said bit signals of said image memory, said means responsive to said comparison comprising means for adjusting said probing in order to probe correct pixels within said read area.

49. An apparatus according to claim 48, wherein said means responsive to said comparison comprising means for adjusting said probing prior to said entire document being sensed.

50. An apparatus according to claim 48, wherein aid document reference characteristics comprise edges of said document.

51. An apparatus according to claim 50, wherein:
said master document lines include horizontal lines running parallel with said rows of aid bit signals;
said means for calibrating comprising means for measuring said horizontal lines with respect to a leading edge of said document; and said means responsive to said comparison comprising means for providing said probe adjustment along the columns of said image memory.

52. An apparatus according to claim 50, wherein:
said master document lines include vertical lines running parallel with said columns of aid bit signals, wherein said means for calibrating comprises means for measuring said vertical lines measured with respect to a single edge of said document; and said means responsive to said comparison comprising means for providing probe adjustment along the rows of said image memory.

53. An apparatus for reading information on a transaction document having image data thereon, said image data including at least one reading area, comprising:
a linear array sensor for sensing a plurality of rows of image data of said document as it is transported across said sensor, each row including a plurality of dark or light pixel areas, said pixel areas of said rows being aligned in a plurality of columns;
means for generating bit signals representative of the dark or light pixel areas of said image data and for storing said bit signals in an image memory organized into rows and columns of bit signals corresponding to said rows of pixel areas of said image data of said document;
means for locating document reference characteristics by probing said image memory and for determining the angle of said document with respect to a line established by said linear array sensor; and means for compensating for said angle of said document as it is transported across said sensor when the read area is being located and read by probing the bit signals of said image memory, said means for compensating comprising means for adjusting said probing in order to probe correct pixels within the read area.

54. An apparatus according to claim 53, further comprising transport means for transporting said document across said sensor, said transport means including a document sensor for sensing whether said document is along an input edge of said transport means prior to initializing drive roller means, said drive roller means comprising means for driving said document against said input edge as it transport said document across said linear array sensor.

55. Apparatus according to claim 53, wherein said document reference characteristics include a side edge and a leading edge transverse to said edge.

56. Apparatus according to claim 53, wherein said document reference characteristics include a timing mark on said document.

57. A transaction method, comprising the steps of:
providing a user with a user identification card having an identification number thereon;
providing the user with a user-personalized document;
digitally imaging said user identification card number and said user-parsonalized document as they are transported through a document reader;
storing said digital information representing said user identification number and associated user-personalized document;
completing a document transaction by transporting a transaction document and said user identification card with said user's identification card number thereon through said document reader;
optically recognizing said identification number of said identification card and reading said transaction document; and associating said identification number with said transaction document, so that said personalized document is not required with each transaction.

58. A transaction method according to claim 57 wherein said user identification card is provided to the user with said identification number encoded on a magnetic strip thereon.

59. A transaction method according to claim 57 wherein said user-personalized document is characterized with said user's name and address when it is digitally imaged.

60. A transaction method according to claim 57 wherein said user-personalized document is characterized with an account number for said user when said user-personalized document is digitally imaged.

61. A transaction method according to claim 59 wherein:
said transaction document is configured to comprises a transaction receipt; and when said receipt is verified as a claiming receipt, notification is sent to said user by accessing the user's digitized name and address.

62. An apparatus for reading colored image data of at least two colors on a document, said apparatus comprising:
sensing means for sensing a plurality of pixel areas of image data on said document and for generating output signals representative of said pixel areas;
transform means for receiving said output signals and for generating digital signals representative of the pixel areas as a function of threshold values for said at least two colors;
image memory means comprising an image memory for storing said digital signals, said image memory including at least two color image memory fields, one image memory field for each of said at least two colors; and probing means for interrogating at least one of said color image memory fields in order to read image data of a particular color.

63. An apparatus according to claim 62, wherein said image data includes at least one feature of a first color, said probing means comprising means for interrogating said color image memory field corresponding to said first color in order to read said at least one feature.

64. An apparatus according to claim 62, wherein said sensing means comprises:
a first color light source;
a second color light source, said first and second color light sources alternately illuminating a plurality of rows of pixel areas of said document; and linear array sensing means for sensing reflected light from said document and generating output signals representative of said pixel areas.

65. An apparatus according to claim 64, wherein said transform means comprises:
mean for digitizing said output signals representative of said pixel areas resulting in digitized intensity output signals;

transform memory means for generating and storing digital signals for individual columns of pixel areas along said rows of pixel areas as a function of a threshold value for said first light source and for generating and storing digital signals for individual columns of pixel areas along said rows of pixel areas as a function of a threshold value for said second light source;
means for addressing said transform memory means to select digital signals representative of whether said output signals representative of said pixel areas are above or below said threshold values, said image memory storing said digital signals such said image memory represents a first color image memory field corresponding to said first color light source and a second color image memory field corresponding to said second color light source.

66. An apparatus according to claim 65, wherein said image data includes at least one feature of a first color and at least one feature of a second color, said probing means interrogating said first color image memory field to read said at least one feature of said first color and interrogating said second image memory field to read said at least one feature of said second color.

67. An apparatus according to claim 66, wherein said first and second color threshold value are such that a third colored feature is represented in said image memory in both first and second color image memory fields.

68. An apparatus according to claim 62, wherein said sensing means comprises:
a light source for illuminating a plurality of rows of pixel areas of said document; and a color sensing means for sensing reflected light of different colors from said document and for generating output signals representative of the different colors of said plurality of pixel areas.

69. An apparatus according to claim 68, wherein said transform means comprises:
means for digitizing said output signals representative of said pixel areas of different colors;
transform memory means for generating and storing digital signals for individual columns of pixel areas along said rows of pixel areas as a function of a threshold value for each of said different colors, said transform memory means for storing the digital signals corresponding to each of said different colors; and means for addressing said transform memory means to select digital signals representative of whether said output signals representative of said pixel areas are above or below said corresponding threshold value for each color, said image memory storing said selected digital signals such that said stored digital signals form a plurality of color image memory fields, one for each of said different colors.