CN101246602A - Human body posture reconstruction method based on geometry backbone - Google Patents
Human body posture reconstruction method based on geometry backbone Download PDFInfo
- Publication number
- CN101246602A CN101246602A CNA2008100335103A CN200810033510A CN101246602A CN 101246602 A CN101246602 A CN 101246602A CN A2008100335103 A CNA2008100335103 A CN A2008100335103A CN 200810033510 A CN200810033510 A CN 200810033510A CN 101246602 A CN101246602 A CN 101246602A
- Authority
- CN
- China
- Prior art keywords
- human
- human body
- image
- articulation point
- skeleton
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Granted
Links
Images
Abstract
The present invention provides a reconstruction method of human posture based on the geometrical skeleton, including:determining human skeleton pattern; obtaining 3D human skeleton pattern, pre-processing image; obtaining bianry image of separation of the background and the human; acquiring human geometrical skeleton; obtaining linear geometrical skeleton; labelling articulare of the start frame; bonding the manual labelling articulare on the geometrical skeleton, shortening the initial error; tracing the articulare by light stream method; obtaining the new position by the light stream method; processing error judgement and position correction to the new articulare obtained by the trace with the combined geometrical skeleton; processing line correction calculated by the light stream method; re-constructing the human posture; converting the two-dimensinal coordinate of the articulare to the 3D skeleton pattern by the parameter evaluating method of the proportion orthogonal projection model. The invention improves accuracy of the posture reconstruction, obtaining a high processing efficiency, capable of processing stable and effect human posture reconstruction in the intelligent video monitor system.
Description
Technical field
The invention belongs to the intelligent video monitoring technical field, particularly relate to a kind of human body posture reconstruction method based on geometry backbone.
Background technology
In intelligent video monitoring system, human body posture reconstruction is second stage.By human body posture reconstruction, can determine the attitude of human body in the monitor video, by the human body attitude in the retrieval attitude storehouse, judge the hazard level of this attitude, thereby take certain early warning measure.Human body posture reconstruction is a basic problem of computer vision and pattern-recognition, and it is applied to a lot of fields, as video monitoring, sports analysis, auxiliary clinical treatment diagnosis etc.Carry out effective human body posture reconstruction, function that can the extending video supervisory system makes computing machine can more easily learn, analyze and understand human action and behavior, thereby makes video monitoring system become more intelligent, and important researching value is arranged.
Traditional human motion analysis method generally is divided into two kinds.The one, at each joint part difference machine for casing-in or applying cover to book electric transducer of human body.In the human motion process, sensor can constantly return to computing machine with the position of each joint in the space, and computing machine just can accurately obtain human body at each movable information constantly like this.The 2nd, analyze at image sequence.Image sequence can be that single camera takes, and also can take from each visual angle simultaneously by a plurality of video cameras.
At present, in the human body posture reconstruction algorithm that has proposed, great majority adopt and extract feature from images, and mate with manikin, and algorithm more complicated, efficient are also lower.
Summary of the invention
Technical matters to be solved by this invention provides a kind of human geometry's of utilization skeleton, overcome the limitation of classic method, improve the accuracy of posture reconstruction, obtain high processing efficient, can in intelligent video monitoring system, stablize and effective human body posture reconstruction.
The technical solution adopted for the present invention to solve the technical problems is: a kind of human body posture reconstruction method based on geometry backbone is provided, comprises the following steps:
(1) determines the human skeleton model: regard human body as a rigid body set that connects by articulation point, represent a rigid body, obtain the three-dimensional human skeleton model with a line segment;
(2) image pre-service: frame of video is kept boundary profile information by the fuzzy inner vein of bidirectional filtering, carry out the thresholding processing then and obtain bianry image, by different shape filtering, burn into down-sampling, a series of Mathematical Morphology Method of up-sampling it is further handled again, obtain the bianry image that background is separated with human body;
(3) obtain human geometry's skeleton: carry out thinning processing from the human region that splits, try to achieve the wire geometry backbone;
(4) the first frame articulation point of mark: the articulation point of mark or automatic mark is tied on the geometry backbone by hand, limits the position of initial joint unique point, dwindles initial error;
(5) follow the tracks of articulation point with optical flow method: by the mode of mark or the method that marks automatically, the selected articulation point that need follow the tracks of as parameter, is tried to achieve the reposition of articulation point with the articulation point set with optical flow method;
(6) the new articulation point that tracking is obtained in conjunction with geometry backbone carries out that mistake is judged and position correction: according to the Human Physiology priori, will far away the makeing mistakes in normal position that depart from that optical flow method calculates a little be pulled back to be bundled on the geometry backbone and proofread and correct;
(7) human body attitude three-dimensional reconstruction: the method for parameter estimation with ratio rectangular projection model is converted to the three-dimensional framework model with articulation point two-dimensional coordinate sequence.
The method of described step (2) is:
1) earlier former two field picture is carried out bidirectional filtering, the fuzzy objective inner vein keeps object edge contour information;
2) filtered image is changed into gray level image;
3) with different threshold values gray level image being carried out thresholding cuts apart;
4) bianry image after thresholding is cut apart carries out morphologic filtering;
5) carry out repeatedly corrosion and expansion process;
6) successively carrying out down-sampling and up-sampling handles;
7) image pixel is distributed carry out statistical study and obtain net result.
Described step (5) is selected OpenCV vision storehouse for use; Described optical flow method is the iteration Lucas-Kanade optical flow tracking algorithm in the image pyramid.
The concrete operations step as shown in Figure 1.
1. human skeleton model
We regard human body as a set by the rigid body of articulation point connection.Two rigid bodies of last underarm that connected by elbow joint as upper limbs are formed, and upper arm is connected by shoulder joint with trunk.We represent a rigid body with a line segment, and human motion is reduced to the motion of human skeleton, have so just obtained a three-dimensional human skeleton model.As shown in Figure 2, this manikin comprises the articulation point of 15 human bodies altogether, 14 sections chain bars.Table 1-1 has provided the ratio of the partes corporis humani position of this method employing.
| Body part | Relative length |
| Height | 1H |
| Head length | 0.182H |
| Shoulder breadth (partly) | 0.120H |
| Upperarm length | 0.185H |
| Stem length | 0.288H |
| Forearm length | 0.190H |
| Hip wide (partly) | 0.095H |
| Thigh length | 0.275H |
| Leg length | 0.285H |
Table 1-1 the partes corporis humani divide ratio
2. image pre-processing method and process
1) earlier former two field picture is carried out bidirectional filtering, the fuzzy objective inner vein keeps object edge contour information.
2) filtered image is changed into gray level image.
3) with different threshold values gray level image being carried out thresholding cuts apart:
Utilize different threshold values can obtain in various degree segmentation effect, this statistical study for the back provides data.
4) bianry image after thresholding is cut apart carries out morphologic filtering:
The noise of image in the morphologic filtering filtering image through producing after the Threshold Segmentation
5) carry out repeatedly corrosion and expansion process:
Corrosion and expansion process are further removed impurity point, only stay maximum human body parts.
6) successively carrying out down-sampling and up-sampling handles:
The operation of down-sampling and up-sampling is original 1/4th by earlier image being narrowed down to, and then amplifies, thereby reaches the purpose of the impurity point that disappears.
7) image pixel is distributed carry out statistical study and obtain net result:
According to interesting areas, choose different threshold values, former two field picture is done above-mentioned processing, can be partitioned into corresponding part, as the body part that clothes covers, the colour of skin part that no clothes covers, pants part of dress and clothes different colours or the like.These parts that split are merged, just can obtain the human body that we want and the bianry image of background separation.The method that merges is exactly, is that black or white probability carry out statistical study in these images to the point of each location of pixels, and the point of this position of final decision is white or black on earth.
3. obtain human geometry's skeleton technique
Utilize morphology knowledge, provide a kind of a kind of algorithm that the morphology refinement is carried out in the two-value zone of practicality below. also be the employed algorithm of this program. establish the known target point and be labeled as 1, background dot is labeled as 0. frontier point and is meant and itself is labeled as 1 and it 8 is communicated with and has at least to be labeled as a point of 0 in neighborhood. and algorithm promptly all carries out following check and operation in one 3 * 3 zone to all frontier points of piece image:
(1) consider with the frontier point to be 8 neighborhoods at center, establishing p1 is central point, and 8 points of its neighborhood are labeled as p2 respectively around central point counterclockwise, p3 ..., p9, if wherein p2 is positioned at the top of p1. during p1=1 (being stain), below 4 conditions satisfy simultaneously, then delete p1 (p1=0):
1. 2≤N (p1)≤6, wherein N (p1) is the number of the non-zero adjoint point of p1;
2. S (p1)=1, wherein S (p1) is with p2, p3, p4 ..., the number of times that the value of these points from 0 to 1 changed when p9 was preface;
3. p2p4p6=0 or S (p1) ≠ 1;
4. p4p6p8=0 or S (p1) ≠ 1.
(2) with (1) step, the condition in only will be 3. changes p2p4p8=0 into, 4. the condition in change into p2p6p8=0. equally when all frontier points are all checked finish after, with all point deletions that satisfies condition.
More than two step operation (1) (2) constitute an iteration. algorithm iterates, until the condition that does not have a little to satisfy again tag delete, at this moment Sheng Xia point is with regard to the skeleton of compositing area. and Fig. 3 has provided the application example of this algorithm. wherein, Fig. 3 (b), (c), (d) being the unsuppressible three kinds of situations of p1. deletion p1 can cut zone in Fig. 3 (b), and deletion p1 can be cut apart the shortening edge among Fig. 3 (c), and 2≤N (p1)≤6 but p1 unsuppressible-suppression satisfy condition among Fig. 3 (d).(a) mark p1 and adjoint point; (b) p1 unsuppressible-suppression situation one; (c) p1 unsuppressible-suppression situation two; (d) p1 unsuppressible-suppression situation three; (e) image before the refinement; (f) result after the refinement
4. mark first frame articulation point, and utilize optical flow method to carry out the articulation point tracking
We have adopted a kind of tracking based on feature to realize tracking to human joint points.It also can adopt the method for automatic mark at first by the mode of mark, comes the selected articulation point that need follow the tracks of, mainly finishes automatic tracking by optical flow method in subsequent frame then.
OpenCV (
The computer vision of increasing income storehouse) the function cvCalcOpticalFlowPyrLK that provides in has realized the sparse iteration version of Lucas-Kanade optical flow computation in the pyramid.It is according to the unique point coordinate on the former frame unique point coordinate Calculation current video frame that provides.Function is sought the coordinate figure with sub-pixel precision.
5. the new articulation point that tracking is obtained in conjunction with geometry backbone is carried out the method for wrong judgement and position correction
We have adopted different tracking processing policies to the articulation point of human body different parts.
At first, we ask angle point to geometry backbone.As shown in Figure 5, the angle point situation that adopts different threshold values to obtain is different, but acral several points, as the crown, palm and sole, and the point of limbs intersection as neck and belly, all are strong angle points, the probability of being obtained is very big.We choose appropriate threshold, make the geometry backbone of each variation can both obtain these above-mentioned unique points.
Then, the articulation point of all above-mentioned 7 unique points (crown, left hand, the right hand, neck, belly, left foot, right crus of diaphragm) being located when every frame is handled is tied on these 7 angle points nearest apart from it.Can obtain tracking effect accurately like this.
For 4 points of shoulder and hip, according to the physiological structure of human body, these are several to be in static relatively state concerning trunk, so these 4 points are done relative to static according to the position of the unique point of neck and belly and moved.
And for ancon and 4 points of knee actuator current amplifier, we judge whether its position takes place unusually, are example with left elbow, if find the length of the distance of left elbow point and left hand point greater than actual forearm, according to projection theory, this is the situation that impossible occur.So, it is unusual that we are judged to be left elbow point occurrence positions, by it is repositioned onto with left hand point distance for forearm length and on geometry backbone certain a bit, reach the purpose of position correction.The difference of considering video sequence two consecutive frames is little, makes the point of forcing binding to depart from this way, can play the effect of error correction, and can be not fully and expectation state disagree.
6. based on the method for parameter estimation of ratio rectangular projection model
Under ratio rectangular projection, the three-dimensional coordinate of an articulation point on the manikin (X, Y, Z) with its projection coordinate in picture frame (u, v) the relation between can be represented with equation 6.1:
From formula 6.1 as can be seen, under the situation of not considering degree of depth Z, the action effect of the ratio rectangular projection in fact just ratio of volume coordinate changes, and wherein scale factor is exactly the parameter s in the following formula, and following method for reconstructing is around how estimating accurately that scale factor s launches.
See the simplest situation earlier, known one section chain pole length is L, and Fig. 6 is its perspective view under ratio rectangular projection.
Two end points (X among Fig. 6
1, Y
1, Z
1) and (X
2, Y
2, Z
2) projection on image is respectively (u
1, v
1) and (u
2, v
2).If the scale factor s of projection model is known, then be easy to just can obtain the relative depth dZ=(Z of two end points
1-Z
2), it is as follows to derive:
L
2＝(X
1-X
2)
2+(Y
1-Y
2)
2+(Z
1-Z
2)
2 (6.2)
Can get by formula (6.1) again:
Therefore have
To sum up (6.2), (6.4) have:
Directly we calculate two values of dZ by (6.5) formula, but can't determine the symbol of dZ, that is to say to have ambiguousness.This ambiguousness of ratio rectangular projection as shown in Figure 7.As can be seen from Fig. 7, to crossing point (X
1, Y
1, Z
1) reference planes, two (X on its both sides
2, Y
2, Z
2) put all projections and fall same point (u in the image
2, v
2).These 2 relative (X
1, Y
1, Z
1) point the dz size identical, opposite in sign.The symbol of dZ need be by two end points (X
1, Y
1, Z
1) and (X
2, Y
2, Z
2) relative position, promptly the position on their relative reference planes is decided.For the human synovial model, from observer's angle, the distance of joint ionization observer distance that can be from image decides the symbol of relative depth dZ between the adjacent segment point.
On the other hand, consider that from the value of dZ the geometric meaning of depth value has determined that the dZ absolute value can only be a rational number,, should have therefore by (6.5) formula:
L
2＝((u
1-u
2)
2+(v
1-v
2)
2)/s
2≥0 (6.6)
So scale factor s should satisfy following formula:
By formula (6.7) we as can be seen, L is known at the chain pole length, the coordinate (u of the subpoint of two end points on image
1, v
1) and (u
2, v
2) under the available situation, can find the minimum scale factor values s that satisfies inequality (6.7)
Min, its substitution formula (6.5) then can be calculated the value of relative depth dZ.
Above-mentioned derivation is easy to just can be generalized to the situation of the multistage chain bar that connects by articulation point, as shown in Figure 8.Three sections chain bars that connect successively in the way, four articulation point (X of its connection
1, Y
1, Z
1), (X
2, Y
2, Z
2), (X
3, Y
3, Z
3) and (X
4, Y
4, Z
4) we are designated as P with them respectively
1, P
2, P
3, P
4The imaging point of these articulation points on image is respectively (u
1, v
1), (u
2, v
2), (u
3, v
3) and (u
4, v
4).With P
2(X
2, Y
2, Z
2) point is for reference point, reference planes are through this point.
Three sections chain bars among Fig. 8 are used formula (6.7) respectively, obtain as the lower inequality group:
Length L at known each section chain bar
1, L
2, L
3, and the coordinate (u of the imaging point of each articulation point on image
1, v
1), (u
2, v
2), (u
3, v
3) and (u
4, v
4) situation under, can try to achieve each self-corresponding minimum scale factor values s respectively
1, s
2, s
3Get wherein maximum scale factor value as common s
Min, s then
MinCan satisfy three inequality of inequality group (6.8) simultaneously.With s
MinSubstitution formula (6.5) then can calculate P
1The relative P of point
2The absolute value dZ of the some degree of depth
1, P
3The relative P of point
2The absolute value dZ of the degree of depth of point
3, P
4The relative P of point
3The absolute value dZ of the degree of depth of point
4Hypothetical reference point P
2The relative depth value of point is 0, behind the symbol that obtains above three relative depths, just can determine P
1, P
3And P
4The relative depth value of three articulation points.And by (6.3) formula, we can be in the hope of the X coordinate and the Y coordinate of each articulation point.Like this, we just estimate to have obtained the relative three-dimensional coordinate of interconnective 4 articulation points among Fig. 5.
The relative estimation of three sections chain bars among Fig. 8 further is generalized to the given whole human body joint skeleton model of Fig. 2, with belly articulation point (sequence number is 0) is reference point, according to the tree structure of manikin, can estimate to obtain the relative three-dimensional coordinate of 16 articulation points in the model equally.To each frame of video sequence, the such constraint condition of other all corresponding above-mentioned inequality of 15 articulation points (6.7) of sequence number 1～15 in the manikin can be calculated the scale factor s of the minimum that satisfies this constraint condition separately thus respectively.And concerning the whole human body model, the only corresponding scale factor of a projection imaging, this scale factor should satisfy the shape that is made of all 15 the inequality constrain conditions inequality group suc as formula (6.8).Therefore,, get maximal value in 15 s values that estimation respectively obtains, realize the initial parameter estimation in the single frames video as the initial estimate s of the projection scale factor of this frame to each frame of video.
Beneficial effect
1) by using bidirectional filtering and a series of morphological transformation that image is carried out pre-service, improves the accuracy that target body is extracted.
2) by obtaining the geometry backbone of target in advance, select Different Strategies discriminatively for use, the articulation point of different parts is positioned, improve the accuracy that articulation point is followed the tracks of, operational efficiency is also higher.
3) employing does not need to carry out camera calibration based on the method for parameter estimation of ratio rectangular projection model, and any video sequence all is suitable for, and has strengthened the applicability of system.
Description of drawings
Fig. 1 is the process flow diagram of human body posture reconstruction method of the present invention.
Fig. 2 is a human skeleton model of the present invention.
Fig. 3 is the synoptic diagram of thinning algorithm of the present invention.
Fig. 4 is the synoptic diagram that the human body that separates is carried out the geometry backbone that refinement obtains from background of the present invention.
Fig. 5 is for trying to achieve the synoptic diagram of angle point under the different threshold value situations of the present invention.
Fig. 6 is a ratio rectangular projection synoptic diagram of the present invention.
Fig. 7 is the synoptic diagram of the ambiguousness in the ratio rectangular projection of the present invention.
Fig. 8 is the ratio rectangular projection synoptic diagram of three sections chain bars of the present invention.
Fig. 9 is an experiment effect synoptic diagram of the present invention.
Embodiment
Below in conjunction with specific embodiment, further set forth the present invention.Should be understood that these embodiment only to be used to the present invention is described and be not used in and limit the scope of the invention.Should be understood that in addition those skilled in the art can make various changes or modifications the present invention after the content of having read the present invention's instruction, these equivalent form of values fall within the application's appended claims institute restricted portion equally.
Embodiment, after consider carrying out efficient and transplantability, native system uses standard C ++ realize, and by OpenCV (
The computer vision of increasing income storehouse) helps realize.
Corresponding to the method among the present invention,, the title of task and the function of task have been described in the table 2 for the human body posture reconstruction system has designed 6 tasks altogether.
| Task names | Function |
| The image pre-service | Human body is split from the background image of frame of video |
| Ask geometry backbone | Try to achieve geometry backbone from the human region that splits with the method for refinement |
| Mark first frame articulation point | The articulation point of manual mark is tied on the geometry backbone, reduces initial error |
| Follow the tracks of articulation point with optical flow method | Calculate the reposition of the articulation point of former frame with optical flow method |
| Judge and correction based on the error of human body priori | The new coordinate of the point that optical flow method calculates is not necessarily correct, needs according to the Human Physiology priori it to be proofreaied and correct. |
| The human body attitude three-dimensional reconstruction | Use OpenGL,, articulation point two-dimensional coordinate sequence is converted to the three-dimensional framework model in conjunction with method for parameter estimation based on ratio rectangular projection model. |
The explanation of table 2 human body posture reconstruction system task
Referring to Fig. 1, whole human body posture reconstruction implementation procedure is as follows: by the image pre-service, obtain isolated human region from background earlier; According to segmentation result, find the solution the geometry backbone of human body with the method for morphology refinement then; Then the joint unique point of mark human body on first frame of video sequence can limit the position of initial joint unique point according to existing geometry backbone at this moment, dwindles initial error; As parameter, the optical flow method function among the input OpenCV is followed the tracks of the motion of human body, tries to achieve the reposition of articulation point with the articulation point set; May there be the error of " obviously " in the new coordinate of the articulation point that optical flow method is tried to achieve, the projected length that is certain section bone is also longer than physical length, this situation, we will depart from far away the makeing mistakes in normal position and a little pull back and be bundled on the geometry backbone, promptly error be judged and be proofreaied and correct; To the new articulation point after proofreading and correct, the method for parameter estimation of utilization ratio rectangular projection model is converted to the three-dimensional framework model with articulation point two-dimensional coordinate sequence, thereby finishes the three-dimensional reconstruction of human body attitude.
Claims (3)
1. the human body posture reconstruction method based on geometry backbone comprises the following steps:
(1) determines the human skeleton model: regard human body as a rigid body set that connects by articulation point, represent a rigid body, obtain the three-dimensional human skeleton model with a line segment;
(2) image pre-service: frame of video is kept boundary profile information by the fuzzy inner vein of bidirectional filtering, carry out the thresholding processing then and obtain bianry image, by different shape filtering, burn into down-sampling, a series of Mathematical Morphology Method of up-sampling it is further handled again, obtain the bianry image that background is separated with human body;
(3) obtain human geometry's skeleton: carry out thinning processing from the human region that splits, try to achieve the wire geometry backbone;
(4) the first frame articulation point of mark: the articulation point of mark or automatic mark is tied on the geometry backbone by hand, limits the position of initial joint unique point, dwindles initial error;
(5) follow the tracks of articulation point with optical flow method: by the mode of mark or the method that marks automatically, the selected articulation point that need follow the tracks of as parameter, is tried to achieve the reposition of articulation point with the articulation point set with optical flow method;
(6) the new articulation point that tracking is obtained in conjunction with geometry backbone carries out that mistake is judged and position correction: according to the Human Physiology priori, will far away the makeing mistakes in normal position that depart from that optical flow method calculates a little be pulled back to be bundled on the geometry backbone and proofread and correct;
(7) human body attitude three-dimensional reconstruction: the method for parameter estimation with ratio rectangular projection model is converted to the three-dimensional framework model with articulation point two-dimensional coordinate sequence.
2. a kind of human body posture reconstruction method based on geometry backbone according to claim 1 is characterized in that: the method for described step (2) is:
1) earlier former two field picture is carried out bidirectional filtering, the fuzzy objective inner vein keeps object edge contour information;
2) filtered image is changed into gray level image;
3) with different threshold values gray level image being carried out thresholding cuts apart;
4) bianry image after thresholding is cut apart carries out morphologic filtering;
5) carry out repeatedly corrosion and expansion process;
6) successively carrying out down-sampling and up-sampling handles;
7) image pixel is distributed carry out statistical study and obtain net result.
3. a kind of human body posture reconstruction method based on geometry backbone according to claim 1 is characterized in that: described step (5) is selected OpenCV vision storehouse for use; Described optical flow method is the iteration Lucas-Kanade optical flow tracking algorithm in the image pyramid.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN2008100335103A CN101246602B (en) | 2008-02-04 | 2008-02-04 | Human body posture reconstruction method based on geometry backbone |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN2008100335103A CN101246602B (en) | 2008-02-04 | 2008-02-04 | Human body posture reconstruction method based on geometry backbone |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN101246602A true CN101246602A (en) | 2008-08-20 |
| CN101246602B CN101246602B (en) | 2010-09-08 |
Family
ID=39947038
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN2008100335103A Expired - Fee Related CN101246602B (en) | 2008-02-04 | 2008-02-04 | Human body posture reconstruction method based on geometry backbone |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN101246602B (en) |
Cited By (42)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN101840584A (en) * | 2010-04-23 | 2010-09-22 | 浙江工商大学 | Two-dimensional digital character operating method based on interactive display |
| CN101840508A (en) * | 2010-04-26 | 2010-09-22 | 中国科学院计算技术研究所 | Method and system for automatically identifying characteristic points in human body chain structure. |
| CN101894377A (en) * | 2010-06-07 | 2010-11-24 | 中国科学院计算技术研究所 | Tracking method of three-dimensional mark point sequence and system thereof |
| CN101907459A (en) * | 2010-07-12 | 2010-12-08 | 清华大学 | Monocular video based real-time posture estimation and distance measurement method for three-dimensional rigid body object |
| CN101958008A (en) * | 2010-10-12 | 2011-01-26 | 上海交通大学 | Automatic texture mapping method in three-dimensional reconstruction of sequence image |
| CN102073909A (en) * | 2010-12-28 | 2011-05-25 | 成都鹏业软件股份有限公司 | Implementation method for high-efficiency semiautomatic artificial intelligence (AI) software |
| CZ302578B6 (en) * | 2010-03-26 | 2011-07-20 | Hanzelka@Tomáš | Method of increasing accuracy of 3D X-ray image |
| CZ302594B6 (en) * | 2010-04-09 | 2011-07-27 | Univerzita Tomáše Bati ve Zlíne | Replication method of surface structures |
| CN102157009A (en) * | 2011-05-24 | 2011-08-17 | 中国科学院自动化研究所 | Method for compiling three-dimensional human skeleton motion based on motion capture data |
| CN102339477A (en) * | 2010-07-16 | 2012-02-01 | 中国科学院计算技术研究所 | Method for automatically establishing human body chain structure model and system thereof |
| CN102541258A (en) * | 2010-11-19 | 2012-07-04 | 微软公司 | Bi-modal depth-image analysis |
| CN102591459A (en) * | 2010-12-28 | 2012-07-18 | 微软公司 | Classification of posture states |
| CN102778953A (en) * | 2012-06-28 | 2012-11-14 | 华东师范大学 | Motion sensing control method of shadow play remote digital performing based on Kinect |
| CN103268629A (en) * | 2013-06-03 | 2013-08-28 | 程志全 | Mark-point-free real-time restoration method of three-dimensional human form and gesture |
| CN101789125B (en) * | 2010-01-26 | 2013-10-30 | 北京航空航天大学 | Method for tracking human skeleton motion in unmarked monocular video |
| CN103440037A (en) * | 2013-08-21 | 2013-12-11 | 中国人民解放军第二炮兵工程大学 | Real-time interaction virtual human body motion control method based on limited input information |
| CN103471500A (en) * | 2013-06-05 | 2013-12-25 | 江南大学 | Conversion method of plane coordinate and space three-dimensional coordinate point in vision of monocular machine |
| CN103679175A (en) * | 2013-12-13 | 2014-03-26 | 电子科技大学 | Fast 3D skeleton model detecting method based on depth camera |
| CN103824326A (en) * | 2014-03-05 | 2014-05-28 | 北京工业大学 | Dynamic human body three-dimensional modeling method |
| CN103914869A (en) * | 2014-02-26 | 2014-07-09 | 浙江工业大学 | Light-weight three-dimensional tree model building method supporting skeleton personalization edition |
| CN104463945A (en) * | 2013-11-25 | 2015-03-25 | 安徽寰智信息科技股份有限公司 | Human body posture rebuilding method |
| CN104460971A (en) * | 2013-11-25 | 2015-03-25 | 安徽寰智信息科技股份有限公司 | Human motion rapid capturing method |
| CN104778661A (en) * | 2014-01-10 | 2015-07-15 | 北京三星通信技术研究有限公司 | Skeleton smoothing method and skeleton smoothing equipment used in pose estimation |
| CN104834913A (en) * | 2015-05-14 | 2015-08-12 | 中国人民解放军理工大学 | Flag signal identification method and apparatus based on depth image |
| CN104899921A (en) * | 2015-06-04 | 2015-09-09 | 杭州电子科技大学 | Single-view video human body posture recovery method based on multi-mode self-coding model |
| US9171264B2 (en) | 2010-12-15 | 2015-10-27 | Microsoft Technology Licensing, Llc | Parallel processing machine learning decision tree training |
| CN105184861A (en) * | 2015-10-20 | 2015-12-23 | 南京大学 | Automatic human three-dimensional (3D) reconstruction system |
| CN103999126B (en) * | 2011-12-16 | 2017-02-15 | 吉士特贡有限公司 | Method and device for estimating a pose |
| CN107220596A (en) * | 2017-05-11 | 2017-09-29 | 西安电子科技大学 | Estimation method of human posture based on cascade mechanism for correcting errors |
| CN108074219A (en) * | 2016-11-11 | 2018-05-25 | 上海东软医疗科技有限公司 | A kind of method for correcting image, device and Medical Devices |
| CN108510582A (en) * | 2018-04-02 | 2018-09-07 | 苏州蜗牛数字科技股份有限公司 | A kind of rigid body generation method of the arbitrary skeleton model of matching |
| CN108537284A (en) * | 2018-04-13 | 2018-09-14 | 东莞松山湖国际机器人研究院有限公司 | Posture assessment scoring method based on computer vision deep learning algorithm and system |
| CN108564618A (en) * | 2018-04-11 | 2018-09-21 | 清华大学 | Hand geometry motion method for reconstructing and device based on multi-voxel proton block |
| CN108629801A (en) * | 2018-05-14 | 2018-10-09 | 华南理工大学 | A kind of three-dimensional (3 D) manikin posture of video sequence and Shape Reconstruction method |
| CN109886194A (en) * | 2019-02-21 | 2019-06-14 | 湖北汽车工业学院 | A kind of processing method and equipment of action recognition |
| CN110135375A (en) * | 2019-05-20 | 2019-08-16 | 中国科学院宁波材料技术与工程研究所 | More people's Attitude estimation methods based on global information integration |
| CN110148217A (en) * | 2019-05-24 | 2019-08-20 | 北京华捷艾米科技有限公司 | A kind of real-time three-dimensional method for reconstructing, device and equipment |
| CN110415336A (en) * | 2019-07-12 | 2019-11-05 | 清华大学 | High-precision human posture method for reconstructing and system |
| US10671841B2 (en) | 2011-05-02 | 2020-06-02 | Microsoft Technology Licensing, Llc | Attribute state classification |
| CN112560637A (en) * | 2020-12-10 | 2021-03-26 | 长沙理工大学 | Clothing analysis method, equipment and storage medium based on deep learning |
| CN113487674A (en) * | 2021-07-12 | 2021-10-08 | 北京未来天远科技开发有限公司 | Human body pose estimation system and method |
| TWI797048B (en) * | 2022-09-01 | 2023-03-21 | 洞見資訊股份有限公司 | Skeleton correction equipment design system and skeleton correction equipment thereof |
Families Citing this family (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN108965850B (en) * | 2018-07-05 | 2020-04-07 | 盎锐（上海）信息科技有限公司 | Human body shape acquisition device and method |
-
2008
- 2008-02-04 CN CN2008100335103A patent/CN101246602B/en not_active Expired - Fee Related
Cited By (64)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN101789125B (en) * | 2010-01-26 | 2013-10-30 | 北京航空航天大学 | Method for tracking human skeleton motion in unmarked monocular video |
| CZ302578B6 (en) * | 2010-03-26 | 2011-07-20 | Hanzelka@Tomáš | Method of increasing accuracy of 3D X-ray image |
| CZ302594B6 (en) * | 2010-04-09 | 2011-07-27 | Univerzita Tomáše Bati ve Zlíne | Replication method of surface structures |
| CN101840584A (en) * | 2010-04-23 | 2010-09-22 | 浙江工商大学 | Two-dimensional digital character operating method based on interactive display |
| CN101840584B (en) * | 2010-04-23 | 2012-05-23 | 浙江工商大学 | Two-dimensional digital character operating method based on interactive display |
| CN101840508A (en) * | 2010-04-26 | 2010-09-22 | 中国科学院计算技术研究所 | Method and system for automatically identifying characteristic points in human body chain structure. |
| CN101840508B (en) * | 2010-04-26 | 2013-01-09 | 中国科学院计算技术研究所 | Method and system for automatically identifying characteristic points in human body chain structure. |
| CN101894377A (en) * | 2010-06-07 | 2010-11-24 | 中国科学院计算技术研究所 | Tracking method of three-dimensional mark point sequence and system thereof |
| CN101907459A (en) * | 2010-07-12 | 2010-12-08 | 清华大学 | Monocular video based real-time posture estimation and distance measurement method for three-dimensional rigid body object |
| CN102339477B (en) * | 2010-07-16 | 2013-08-21 | 中国科学院计算技术研究所 | Method for automatically establishing human body chain structure model and system thereof |
| CN102339477A (en) * | 2010-07-16 | 2012-02-01 | 中国科学院计算技术研究所 | Method for automatically establishing human body chain structure model and system thereof |
| CN101958008A (en) * | 2010-10-12 | 2011-01-26 | 上海交通大学 | Automatic texture mapping method in three-dimensional reconstruction of sequence image |
| CN102541258B (en) * | 2010-11-19 | 2015-12-09 | 微软技术许可有限责任公司 | Bi-modal depth-image analysis |
| CN102541258A (en) * | 2010-11-19 | 2012-07-04 | 微软公司 | Bi-modal depth-image analysis |
| US9349040B2 (en) | 2010-11-19 | 2016-05-24 | Microsoft Technology Licensing, Llc | Bi-modal depth-image analysis |
| US9171264B2 (en) | 2010-12-15 | 2015-10-27 | Microsoft Technology Licensing, Llc | Parallel processing machine learning decision tree training |
| CN102073909B (en) * | 2010-12-28 | 2013-07-17 | 成都鹏业软件股份有限公司 | Implementation method for high-efficiency semiautomatic artificial intelligence (AI) software |
| CN102591459A (en) * | 2010-12-28 | 2012-07-18 | 微软公司 | Classification of posture states |
| CN102073909A (en) * | 2010-12-28 | 2011-05-25 | 成都鹏业软件股份有限公司 | Implementation method for high-efficiency semiautomatic artificial intelligence (AI) software |
| CN102591459B (en) * | 2010-12-28 | 2014-10-29 | 微软公司 | Classification of posture states |
| US10671841B2 (en) | 2011-05-02 | 2020-06-02 | Microsoft Technology Licensing, Llc | Attribute state classification |
| CN102157009A (en) * | 2011-05-24 | 2011-08-17 | 中国科学院自动化研究所 | Method for compiling three-dimensional human skeleton motion based on motion capture data |
| CN103999126B (en) * | 2011-12-16 | 2017-02-15 | 吉士特贡有限公司 | Method and device for estimating a pose |
| CN102778953B (en) * | 2012-06-28 | 2015-06-24 | 华东师范大学 | Motion sensing control method of shadow play remote digital performing based on Kinect |
| CN102778953A (en) * | 2012-06-28 | 2012-11-14 | 华东师范大学 | Motion sensing control method of shadow play remote digital performing based on Kinect |
| CN103268629B (en) * | 2013-06-03 | 2015-11-11 | 程志全 | Unmarked some real time restoration method of 3 D human body form and attitude |
| CN103268629A (en) * | 2013-06-03 | 2013-08-28 | 程志全 | Mark-point-free real-time restoration method of three-dimensional human form and gesture |
| CN103471500B (en) * | 2013-06-05 | 2016-09-21 | 江南大学 | A kind of monocular camera machine vision midplane coordinate and the conversion method of 3 d space coordinate point |
| CN103471500A (en) * | 2013-06-05 | 2013-12-25 | 江南大学 | Conversion method of plane coordinate and space three-dimensional coordinate point in vision of monocular machine |
| CN103440037B (en) * | 2013-08-21 | 2017-02-08 | 中国人民解放军第二炮兵工程大学 | Real-time interaction virtual human body motion control method based on limited input information |
| CN103440037A (en) * | 2013-08-21 | 2013-12-11 | 中国人民解放军第二炮兵工程大学 | Real-time interaction virtual human body motion control method based on limited input information |
| CN104460971A (en) * | 2013-11-25 | 2015-03-25 | 安徽寰智信息科技股份有限公司 | Human motion rapid capturing method |
| CN104463945A (en) * | 2013-11-25 | 2015-03-25 | 安徽寰智信息科技股份有限公司 | Human body posture rebuilding method |
| CN103679175A (en) * | 2013-12-13 | 2014-03-26 | 电子科技大学 | Fast 3D skeleton model detecting method based on depth camera |
| CN103679175B (en) * | 2013-12-13 | 2017-02-15 | 电子科技大学 | Fast 3D skeleton model detecting method based on depth camera |
| CN104778661A (en) * | 2014-01-10 | 2015-07-15 | 北京三星通信技术研究有限公司 | Skeleton smoothing method and skeleton smoothing equipment used in pose estimation |
| CN103914869A (en) * | 2014-02-26 | 2014-07-09 | 浙江工业大学 | Light-weight three-dimensional tree model building method supporting skeleton personalization edition |
| CN103914869B (en) * | 2014-02-26 | 2017-02-22 | 浙江工业大学 | Light-weight three-dimensional tree model building method supporting skeleton personalization edition |
| CN103824326A (en) * | 2014-03-05 | 2014-05-28 | 北京工业大学 | Dynamic human body three-dimensional modeling method |
| CN104834913B (en) * | 2015-05-14 | 2018-04-03 | 中国人民解放军理工大学 | Semaphore recognition methods and device based on depth image |
| CN104834913A (en) * | 2015-05-14 | 2015-08-12 | 中国人民解放军理工大学 | Flag signal identification method and apparatus based on depth image |
| CN104899921B (en) * | 2015-06-04 | 2017-12-22 | 杭州电子科技大学 | Single-view videos human body attitude restoration methods based on multi-modal own coding model |
| CN104899921A (en) * | 2015-06-04 | 2015-09-09 | 杭州电子科技大学 | Single-view video human body posture recovery method based on multi-mode self-coding model |
| CN105184861B (en) * | 2015-10-20 | 2019-02-12 | 南京大学 | A kind of automatic body three-dimensional reconstruction system |
| CN105184861A (en) * | 2015-10-20 | 2015-12-23 | 南京大学 | Automatic human three-dimensional (3D) reconstruction system |
| CN108074219A (en) * | 2016-11-11 | 2018-05-25 | 上海东软医疗科技有限公司 | A kind of method for correcting image, device and Medical Devices |
| CN107220596A (en) * | 2017-05-11 | 2017-09-29 | 西安电子科技大学 | Estimation method of human posture based on cascade mechanism for correcting errors |
| CN107220596B (en) * | 2017-05-11 | 2020-04-21 | 西安电子科技大学 | Human body posture estimation method based on cascade error correction mechanism |
| CN108510582A (en) * | 2018-04-02 | 2018-09-07 | 苏州蜗牛数字科技股份有限公司 | A kind of rigid body generation method of the arbitrary skeleton model of matching |
| CN108564618B (en) * | 2018-04-11 | 2020-12-01 | 清华大学 | Hand geometric motion reconstruction method and device based on multiple voxel blocks |
| CN108564618A (en) * | 2018-04-11 | 2018-09-21 | 清华大学 | Hand geometry motion method for reconstructing and device based on multi-voxel proton block |
| CN108537284A (en) * | 2018-04-13 | 2018-09-14 | 东莞松山湖国际机器人研究院有限公司 | Posture assessment scoring method based on computer vision deep learning algorithm and system |
| CN108629801B (en) * | 2018-05-14 | 2020-11-24 | 华南理工大学 | Three-dimensional human body model posture and shape reconstruction method of video sequence |
| CN108629801A (en) * | 2018-05-14 | 2018-10-09 | 华南理工大学 | A kind of three-dimensional (3 D) manikin posture of video sequence and Shape Reconstruction method |
| CN109886194A (en) * | 2019-02-21 | 2019-06-14 | 湖北汽车工业学院 | A kind of processing method and equipment of action recognition |
| CN110135375B (en) * | 2019-05-20 | 2021-06-01 | 中国科学院宁波材料技术与工程研究所 | Multi-person attitude estimation method based on global information integration |
| CN110135375A (en) * | 2019-05-20 | 2019-08-16 | 中国科学院宁波材料技术与工程研究所 | More people's Attitude estimation methods based on global information integration |
| CN110148217A (en) * | 2019-05-24 | 2019-08-20 | 北京华捷艾米科技有限公司 | A kind of real-time three-dimensional method for reconstructing, device and equipment |
| CN110415336A (en) * | 2019-07-12 | 2019-11-05 | 清华大学 | High-precision human posture method for reconstructing and system |
| CN112560637A (en) * | 2020-12-10 | 2021-03-26 | 长沙理工大学 | Clothing analysis method, equipment and storage medium based on deep learning |
| CN112560637B (en) * | 2020-12-10 | 2024-03-15 | 长沙理工大学 | Deep learning-based clothing analysis method, equipment and storage medium |
| CN113487674A (en) * | 2021-07-12 | 2021-10-08 | 北京未来天远科技开发有限公司 | Human body pose estimation system and method |
| CN113487674B (en) * | 2021-07-12 | 2024-03-08 | 未来元宇数字科技(北京)有限公司 | Human body pose estimation system and method |
| TWI797048B (en) * | 2022-09-01 | 2023-03-21 | 洞見資訊股份有限公司 | Skeleton correction equipment design system and skeleton correction equipment thereof |
Also Published As
| Publication number | Publication date |
|---|---|
| CN101246602B (en) | 2010-09-08 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN101246602B (en) | Human body posture reconstruction method based on geometry backbone | |
| CN110135375B (en) | Multi-person attitude estimation method based on global information integration | |
| CN105741252B (en) | Video image grade reconstruction method based on rarefaction representation and dictionary learning | |
| CN105488758B (en) | A kind of image-scaling method based on perception of content | |
| CN104794722A (en) | Dressed human body three-dimensional bare body model calculation method through single Kinect | |
| CN110222665A (en) | Human motion recognition method in a kind of monitoring based on deep learning and Attitude estimation | |
| CN104715493A (en) | Moving body posture estimating method | |
| CN102420985B (en) | Multi-view video object extraction method | |
| CN104834893A (en) | Front-view pedestrian gait period detection method | |
| CN101894377A (en) | Tracking method of three-dimensional mark point sequence and system thereof | |
| CN109829972B (en) | Three-dimensional human standard skeleton extraction method for continuous frame point cloud | |
| JP2019096113A (en) | Processing device, method and program relating to keypoint data | |
| CN105069751A (en) | Depth image missing data interpolation method | |
| CN111985579A (en) | Double-person diving synchronism analysis method based on camera cooperation and three-dimensional skeleton estimation | |
| JP2018147313A (en) | Object attitude estimating method, program and device | |
| CN101727568B (en) | Foreground action estimation device and foreground action estimation method | |
| GB2503331A (en) | Aligning garment image with image of a person, locating an object in an image and searching for an image containing an object | |
| CN111311751A (en) | Three-dimensional clothes model reconstruction method based on deep neural network | |
| CN103577800A (en) | Method for measuring human hand morphological parameters based on color images | |
| Kuang et al. | An effective skeleton extraction method based on Kinect depth image | |
| CN112634411B (en) | Animation generation method, system and readable medium thereof | |
| Lin et al. | Automatic feature extraction from front and side images | |
| CN110309554B (en) | Video human body three-dimensional reconstruction method and device based on garment modeling and simulation | |
| CN104132650A (en) | Method for detection of fruit tree leaf density by maximum contour rectangle technique | |
| JPH0273471A (en) | Estimating method for three-dimensional form |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| C06 | Publication | ||
| PB01 | Publication | ||
| C10 | Entry into substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| C14 | Grant of patent or utility model | ||
| GR01 | Patent grant | ||
| C17 | Cessation of patent right | ||
| CF01 | Termination of patent right due to non-payment of annual fee |
Granted publication date: 20100908 Termination date: 20130204 |