CN102663774A - Motion tracking method - Google Patents
Motion tracking method Download PDFInfo
- Publication number
- CN102663774A CN102663774A CN2012100872256A CN201210087225A CN102663774A CN 102663774 A CN102663774 A CN 102663774A CN 2012100872256 A CN2012100872256 A CN 2012100872256A CN 201210087225 A CN201210087225 A CN 201210087225A CN 102663774 A CN102663774 A CN 102663774A
- Authority
- CN
- China
- Prior art keywords
- point
- human body
- body distal
- profile
- maximum
- 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.)
- Pending
Links
Images
Landscapes
- Image Analysis (AREA)
Abstract
Provided in the invention is motion tracking method, comprising the following steps: obtaining a human body end point; and by taking continuity of a position of a same characteristic point in terms of time as a tracking basis, selecting a maximum likelihood position of the human body end point at a t-1 time by employing a maximum likelihood method at a t time.
Description
Technical field
The present invention relates to image processing field, particularly a kind of method of many people motion tracking.
Technical background
Motion target tracking is an important research project that produces in the digital video technology development; So-called motion target tracking; Be exactly on the basis of moving object detection, utilize the target effective characteristic, use suitable matching algorithm; In sequence image, seeking the position of the image the most similar with To Template, is exactly to give target localization in brief.In practical application, movement locus that motion target tracking not only can provide target and accurate localizing objects for next step goal behavior analysis and understanding provides reliable Data Source, and can be offered help for moving object detection.
No matter traditional motion tracking method adopts the method for image difference, or other mode, all undesirable for the effect of many people situation.
Summary of the invention
In view of this, for addressing the above problem, the invention provides a kind of motion tracking method, it has realized the method for people from institute motion tracking well through the several characteristic point is followed the trail of.
In order to achieve the above object; The present invention provides a kind of motion tracking method; It may further comprise the steps: obtain the human body distal point;, adopt maximum-likelihood method to select when t, to engrave the human body distal point when t-1, to engrave the maximum position of possibility as following the trail of foundation with the position continuity in time of same unique point.
Further; Said employing maximum-likelihood method is selected and when t, is engraved the human body distal point engraves the maximum position of possibility when t-1 step and further comprise; Calculate the probability that human body distal point z occurs at t-1 all human body distal points at t constantly constantly, the maximum point of probability is exactly that the human body distal point z that is selected engraves the maximum position of possibility when t-1.
Further, said method further comprises, when calculating said probability with the geodesic distance of said unique point as factor of influence.
Embodiment provided by the invention through calculate tracking point front and back constantly in probability in picture, and introduce geodesic distance as probability factor, can improve the accuracy of tracking greatly, realized many people's motion tracking.
Description of drawings
Fig. 1 is the method flow diagram of a specific embodiment of the present invention.
Fig. 2 blocks the layering synoptic diagram for monomer individual among concrete embodiment of the present invention.
The synoptic diagram of Fig. 3 for putting on the traversal profile among concrete embodiment of the present invention.
Embodiment
This law provides a kind of motion tracking method, and it may further comprise the steps.
In a concrete embodiment, four limbs and head that the human body distal point is behaved, the acquisition methods of its position specifically may further comprise the steps.
Step 1011 is calculated the profile center of gravity of said each monomer individual profile.
In a concrete embodiment, the center of gravity of calculating monomer individual profile adopts the arithmetic mean that calculates every bit coordinate figure on the said profile as the individual profile center of gravity of said monomer.For the monomer that has no to block individual, only need to gather that each point coordinate value get final product on the monomer individual profile, and when a monomer philtrum has the situation of blocking, the part of front and back different depth will inevitably occur.Adopt following method to solve for this situation the present invention.The center of gravity of calculating monomer individual profile adopts the arithmetic mean that calculates every bit coordinate figure on the said profile as the individual profile center of gravity of said monomer.For the monomer that has no to block individual, only need to gather that each point coordinate value get final product on the monomer individual profile, and when a monomer philtrum has the situation of blocking, the part of front and back different depth will inevitably occur.Adopt following method to solve for this situation the present invention.
Step 10111 is divided into different layers for a monomer discontinuous part of the philtrum degree of depth according to depth of field order.
Step 10112 adopts se ed filling algorithm that every layer of filling is complete.
Step 10113 obtains the boundary line between the adjacent two layers.
Step 10114, when point on the profile and profile center of gravity adhered to different layers separately, every bit must pass through the boundary line between the said adjacent two layers to the route of the geodesic distance of profile center of gravity on the calculating profile.
In a concrete embodiment, please referring to shown in Figure 2, personnel's left arm has blocked health forward among the figure, and left arm is defined as ground floor, and trunk is defined as the second layer.Adopt se ed filling algorithm with the second layer be blocked partially filled complete.And with between two-layer because block the breakpoint point A of discontinuous two individuals outlines and put line between the B as the boundary line between the adjacent two layers., calculating need pass through the line of an A and some B when being positioned at point on the ground floor to the geodesic distance of profile center of gravity.
Step 1012 is calculated on said each monomer individual profile every bit to the geodesic distance of profile center of gravity.
In a concrete embodiment, the point on the calculating profile to the method for the geodesic distance of profile center of gravity is:
Step 10121 is cut into slices to the individual of the monomer after layering profile, said section is created as one is communicated with tree, and the root node of said connection tree is the profile center of gravity of monomer individual profile, and its child node is a point on the contiguous slices.
Step 10122 is according to the point on the traversal profile with the degree of depth, until each point of whole profile can both be by said connection tree traversal.
In a concrete embodiment, as shown in Figure 3,3 some a are arranged on the fragment 1 and put b, fragment 2 has a c on the fragment 3.Its mid point a, b are at the second layer, and some c is at ground floor.With the degree of depth when being had on the profile according to traversal, traversal and the point of profile center of gravity (trunk intermediate hollow round dot place) at one deck earlier are again according to other layers of degree of depth traversal.
Traversal is during fragment 1, and some a be owing to can't be communicated to the profile center of gravity, so geodesic distance is infinite, and therefore some b can directly obtain the geodesic distance of a b owing to can directly be communicated to the profile center of gravity;
Upwards and downwards travel through the point on other fragments respectively along the profile center of gravity, when traversing fragment 2,, upgrade the geodesic distance of an a at this moment because some a can be communicated to the profile center of gravity;
When traversing fragment 3, some c is because at ground floor, not with profile center of gravity layer together, the geodesic distance of therefore putting c is at this moment for infinite;
After all fragments of the second layer were all traveled through, continue the traversal ground floor again, the calculating of the geodesic distance of this time point c need be through boundary line between the adjacent two layers.
What present embodiment adopted is slices across, adopts other slicing modes also can reach same effect certainly.Preferential traversal and profile center of gravity can obtain the geodesic distance put on most profiles sooner with the point of layer, if but adopt to travel through in proper order equally according to other and can realize identical effect.
Step 1013 is chosen maximum value as each monomer individual's the four limbs and the location point of head from the function set that geodesic distance is formed.
The point of choosing successively on the profile with a definite sequence is an independent variable; Is that functional value is set up function set with the point on the profile to the geodesic distance of profile center of gravity; Can conveniently find flex point to said function set differentiate, and derivative is that 0 point is the position of four limbs and head after the differentiate.This shows, should the choosing successively of the point on the profile, otherwise unnecessary function flex point will appear.
Obtained just can following the tracks of after the human body distal point of computing to all distal points.In the process of following the trail of a trick unique point, the position of same unique point continuity in time can be used as the foundation of tracking.
Use bayes rule, a posterior probability can be write as the form that three factors multiply each other, such as:
Normalized component
can remove, because he does not influence sorting result.The first factor
represents
at the pixel is
class probability.
It is accurate inadequately only using this factor of human body distal point z.In a concrete embodiment, the calculating that the geodesic distance I of this unique point is participated in probability as one-component can make the result more accurate.Final probability
PCan use following formula to judge.
The present invention be directed to human body terminal position and terminal position corresponding geodesic distance and come tracing movement; Because the individual human body terminal position of each monomer in can fairly simple ground acquisition crowd; As according to depth of field layering etc., therefore can once follow the trail of many people's motion.Because human body terminal position and geodesic distance all are linear calculating, therefore can reach high real-time.
The above is merely preferred embodiment of the present invention, and is in order to restriction the present invention, not all within spirit of the present invention and principle, any modification of being done, is equal to replacement etc., all should be included within protection scope of the present invention.
Claims (3)
1. motion tracking method; It may further comprise the steps: obtain the human body distal point;, adopt maximum-likelihood method to select when t, to engrave the human body distal point when t-1, to engrave the maximum position of possibility as following the trail of foundation with the position continuity in time of same unique point.
2. method according to claim 1; It is characterized in that; Said employing maximum-likelihood method is selected and when t, is engraved the human body distal point engraves the maximum position of possibility when t-1 step and further comprise; Calculate the probability that human body distal point z occurs at t-1 all human body distal points at t constantly constantly, the maximum point of probability is exactly that the human body distal point z that is selected engraves the maximum position of possibility when t-1.
3. method according to claim 2 is characterized in that, said method further comprises, when calculating said probability with the geodesic distance of said unique point as factor of influence.
Priority Applications (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN2012100872256A CN102663774A (en) | 2012-03-29 | 2012-03-29 | Motion tracking method |
Applications Claiming Priority (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN2012100872256A CN102663774A (en) | 2012-03-29 | 2012-03-29 | Motion tracking method |
Publications (1)
Publication Number | Publication Date |
---|---|
CN102663774A true CN102663774A (en) | 2012-09-12 |
Family
ID=46773252
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
CN2012100872256A Pending CN102663774A (en) | 2012-03-29 | 2012-03-29 | Motion tracking method |
Country Status (1)
Country | Link |
---|---|
CN (1) | CN102663774A (en) |
Cited By (1)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN113158953A (en) * | 2021-04-30 | 2021-07-23 | 青岛海信智慧生活科技股份有限公司 | Personnel searching method, device, equipment and medium |
-
2012
- 2012-03-29 CN CN2012100872256A patent/CN102663774A/en active Pending
Cited By (2)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN113158953A (en) * | 2021-04-30 | 2021-07-23 | 青岛海信智慧生活科技股份有限公司 | Personnel searching method, device, equipment and medium |
CN113158953B (en) * | 2021-04-30 | 2022-11-25 | 青岛海信智慧生活科技股份有限公司 | Personnel searching method, device, equipment and medium |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
US11951373B2 (en) | Automated or assisted umpiring of baseball game using computer vision | |
CN104915970B (en) | A kind of multi-object tracking method based on Track association | |
CN104094279B (en) | Large-range-first cross-camera visual target re-identification method | |
CN110163889A (en) | Method for tracking target, target tracker, target following equipment | |
CN105551086A (en) | Customized foot modeling and shoe pad customization method on the basis of computer vision | |
CN109636829A (en) | A kind of multi-object tracking method based on semantic information and scene information | |
CN102682477B (en) | Regular scene three-dimensional information extracting method based on structure prior | |
CN108279436B (en) | A kind of method and device for picking up seismic first breaks | |
CN107369183A (en) | Towards the MAR Tracing Registration method and system based on figure optimization SLAM | |
CN104599314A (en) | Three-dimensional model reconstruction method and system | |
CN106780484A (en) | Robot interframe position and orientation estimation method based on convolutional neural networks Feature Descriptor | |
KR20110113152A (en) | Apparatus, method and computer-readable medium providing marker-less motion capture of human | |
JP6344953B2 (en) | Trajectory analysis apparatus and trajectory analysis method | |
CN106525000A (en) | A road marking line automatic extracting method based on laser scanning discrete point strength gradients | |
CN104937607A (en) | Device, method, and computer program for reconstructing a motion of an object | |
Morimitsu et al. | Exploring structure for long-term tracking of multiple objects in sports videos | |
CN111062971B (en) | Deep learning multi-mode-based mud head vehicle tracking method crossing cameras | |
CN109063549A (en) | High-resolution based on deep neural network is taken photo by plane video moving object detection method | |
CN107730501A (en) | A kind of fish body tail frequency detection method and system | |
CN110433467A (en) | Picking up table tennis ball robot operation method and equipment based on binocular vision and ant group algorithm | |
KR102369945B1 (en) | Device and method to discriminate excersice stance using pressure | |
Woinoski et al. | Towards automated swimming analytics using deep neural networks | |
CN107341179A (en) | Generation method, device and the storage device of standard movement database | |
Li et al. | Video-based table tennis tracking and trajectory prediction using convolutional neural networks | |
CN102663774A (en) | Motion tracking method |
Legal Events
Date | Code | Title | Description |
---|---|---|---|
C06 | Publication | ||
PB01 | Publication | ||
C02 | Deemed withdrawal of patent application after publication (patent law 2001) | ||
WD01 | Invention patent application deemed withdrawn after publication |
Application publication date: 20120912 |