CN103984916A - Person comparison method based on combined template cluster sampling matching - Google Patents
Person comparison method based on combined template cluster sampling matching Download PDFInfo
- Publication number
- CN103984916A CN103984916A CN201410121866.8A CN201410121866A CN103984916A CN 103984916 A CN103984916 A CN 103984916A CN 201410121866 A CN201410121866 A CN 201410121866A CN 103984916 A CN103984916 A CN 103984916A
- Authority
- CN
- China
- Prior art keywords
- limbs
- personage
- person
- coupling
- comparison method
- 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
- 238000000034 method Methods 0.000 title claims abstract description 32
- 238000005070 sampling Methods 0.000 title claims abstract description 11
- 210000003414 extremity Anatomy 0.000 claims description 65
- 230000008878 coupling Effects 0.000 claims description 24
- 238000010168 coupling process Methods 0.000 claims description 24
- 238000005859 coupling reaction Methods 0.000 claims description 24
- 238000009416 shuttering Methods 0.000 claims description 16
- 239000000284 extract Substances 0.000 claims description 10
- 230000002860 competitive effect Effects 0.000 claims description 7
- 210000001699 lower leg Anatomy 0.000 claims description 6
- 210000000689 upper leg Anatomy 0.000 claims description 6
- 210000000245 forearm Anatomy 0.000 claims description 5
- 239000000203 mixture Substances 0.000 claims description 4
- 230000003213 activating effect Effects 0.000 claims description 3
- 150000001875 compounds Chemical class 0.000 claims description 3
- 238000001514 detection method Methods 0.000 claims description 3
- 238000010586 diagram Methods 0.000 description 2
- 238000012544 monitoring process Methods 0.000 description 2
- 230000007812 deficiency Effects 0.000 description 1
- 238000005516 engineering process Methods 0.000 description 1
- 238000000605 extraction Methods 0.000 description 1
- 238000005286 illumination Methods 0.000 description 1
- 230000000007 visual effect Effects 0.000 description 1
Landscapes
- Measurement Of The Respiration, Hearing Ability, Form, And Blood Characteristics Of Living Organisms (AREA)
- Image Analysis (AREA)
Abstract
The invention discloses a person comparison method based on combined template cluster sampling matching. The method comprises the following steps: 1) a plurality of single-person pictures and one multiple-person picture are inputted, wherein the persons in the plurality of single-person pictures are a same person; 2) a single-person multiple-instance combined template is extracted from the plurality of single-person pictures, and an alternative target set is extracted from the multiple-person picture; 3) a body matching candidate graph is established through the single-person multiple-instance combined template and the alternative target set; 4) and the body matching candidate graph is solved to find the best person body matching. According to the method, the position and logical relationship among person bodies is taken into account, so the accuracy rate of person comparison can be effectively improved.
Description
Technical field
The present invention relates to personage and compare field, more specifically, relate to a kind of personage's comparison method based on gang form cluster sampling coupling.
Background technology
Personage's comparison has started to come into one's own gradually in video monitoring, and particularly in the time that face recognition application is restricted, the function of personage's comparison is just particularly important.But personage's comparison is very difficult and challenge, be difficult to find on the one hand a kind of for personage and the good expression way of robust, because body configuration can vary widely (as visual angle, action, illumination condition etc.), so thereby be difficult to build template and effectively identify personage by extracting rudimentary characteristics of image; Be difficult on the other hand find effective personage's local matching method, a given personage's template, carries out personage's matching result with overall people information and always can obtain a lot of wrong positive samples.Because in real monitor video, the target person that will contrast may be blocked, also may be connected with other personages or background, and at this moment just need to mate and compare by the information of personage part.
Summary of the invention
In order to overcome above-mentioned the deficiencies in the prior art, the invention provides a kind of personage's comparison method based on gang form cluster sampling coupling, the accuracy rate of this comparison method is higher.
Technical scheme of the present invention is:
Based on personage's comparison method of gang form cluster sampling coupling, comprise the following steps:
1) input several single pictures and many people of width picture, wherein single in several single pictures is same people;
2) from several single pictures, extract single many example set shuttering T, and from many people picture, extract alternative target S set;
3) build limbs matching candidate figure G by single many example set shuttering T and alternative target S set;
4) limbs matching candidate figure G is solved, find the coupling of best human limbs.
In the preferred scheme of one, described step 2) from several single pictures, extract single many example set shuttering T, its single many example set shuttering T is:
；
N=6, human limbs is divided into 6 parts, is respectively head, trunk, upper arm, forearm, thigh and shank; Wherein, g represents the concrete human limbs that each is detected, and represents a five-tuple for this rectangle frame with a rectangle frame
express, t represents the type of these limbs, and x and y represent the coordinate at this limbs center, and θ represents the anglec of rotation of these limbs, and s represents the relative scale of these limbs;
The concrete extracting method of above-mentioned single many example set shuttering T is:
21) single picture in the vertical direction is divided into 4 layers, in every one deck, scans detection respectively with corresponding human limbs detecting device, is detected accordingly mark, and concrete layered approach is as follows:
Ground floor: head
The second layer: trunk, upper arm, forearm
The 3rd layer: thigh
The 4th layer: shank
22) reject the limbs that are less than threshold X in testing result with the overlapping ratio of foreground mask, by remaining limbs composition template T.
In the preferred scheme of one, described step 2) extracting method that extracts alternative target S set from many people picture is: scan many people picture with human limbs detecting device, obtain testing result, reject according to foreground mask the limbs that detect that are less than threshold X with prospect overlapping area again, by remaining composition alternative target S set.
In the preferred scheme of one, described threshold X is 75%.
In the preferred scheme of one, described foreground mask is to generate or manually mark generation by the method for background modeling.
In the preferred scheme of one, the coupling that the summit of limbs matching candidate figure G in described step 3) is defined as corresponding limbs in single many example set shuttering T and target alternative S set is right, and the limit of limbs matching candidate figure G is defined as compatibility relation and the competitive relation on adjacent two summits;
Described compatibility relation is to encourage coupling to mutually activating in the process of coupling, and described compatibility relation is expressed as the close degree of two alternative target limb parts, mainly contains two aspects: the kinematic relation that (a) has in position dependence; (b) symmetrical symmetric relation in position;
Described competitive relation is to have suppressed conflict in coupling when being activated simultaneously, mainly contains two aspects: two alternative target that (i) have same limbs type can not be activated simultaneously; (ii) lap of two alternative target should only be compared once.
In the preferred scheme of one, described step 4) is to use compound cluster sampling algorithm to solve limbs matching candidate figure G, finds the coupling of best human limbs.
Compared with prior art, the present invention fully takes into account position and the logical relation between human limbs, so the accuracy rate of personage's comparison is higher.
Brief description of the drawings
Fig. 1 is the inventive method process flow diagram.
Fig. 2 is limbs testing result figure, and wherein (a) is single picture testing result, is (b) many people image detect result.
Fig. 3 is limbs pairing candidate figure schematic diagram.
Fig. 4 is limbs compatibility relation figure, and wherein (a) is kinematic relation (dotted line limit) and the symmetric relation (solid line limit) between limbs, (b) example for combining between limbs.
Embodiment
Below in conjunction with accompanying drawing, the present invention will be further described, but embodiments of the present invention are not limited to this.
A kind of flow process of the personage's comparison method based on gang form cluster sampling coupling as shown in Figure 1, comprises the steps:
1) input several single (same people) pictures and many people of width picture;
Preferably, the present invention follows following two settings in monitoring application demand:
A) personage's dressing remains unchanged in different scenes, and in several single pictures of input and many people of width picture, personage's dressing should remain unchanged;
The personage that b) will compare need to have certain resolution, and it is highly 120 pixels in the present embodiment.
2) from several single pictures, extract single many example set shuttering T, and from many people picture, extract alternative target S set;
Single many example set shuttering extracting method is as follows:
Single picture in the vertical direction is divided into 4 layers, in every one deck, scans detection respectively with corresponding human limbs detecting device, is detected accordingly mark, and concrete layered approach is as follows:
Ground floor: head
The second layer: trunk, upper arm, forearm
The 3rd layer: thigh
The 4th layer: shank
As shown in Figure 2 (a) shows, the head that red wire frame representation detects, the trunk that blue wire frame representation detects, the thigh that yellow wire frame representation detects, the shank that green wire frame representation detects.
Then, then reject the limbs that detect that are less than 75% with prospect overlapping area according to foreground mask, the foreground mask here adopts manually mark to generate.
As shown in Fig. 2 (b), the method of extracting alternative target set from many people picture is similar with the single many case templates of extraction, difference is not divide 4 layers picture in the vertical direction etc., directly with each limbs detecting device scanning full figure sheet, obtain scanning result, and then reject the limbs that detect that are less than 75% with prospect overlapping area according to foreground mask, the foreground mask is here also that manually mark generates.
In the present embodiment, number respectively to the limbs that detect in single picture and many people picture.
3) build limbs matching candidate figure with single many example set shuttering and alternative target set;
The coupling that the summit of limbs matching candidate figure is defined as corresponding limbs in single many example set shuttering and target alternative set is right, as (32,34), (24,24) etc., wherein the former is the limbs numbering in single case template, and the latter is the limbs numbering in alternative set; The limit of limbs matching candidate figure is defined as compatibility relation and the competitive relation on adjacent two summits.As shown in Figure 3, solid line represents compatibility relation, and dotted line represents competitive relation.
Compatibility relation is to encourage coupling to mutually activating in the process of coupling.In the present embodiment, compatibility relation is expressed as to the close degree of two alternative target limb parts, mainly considers two aspects: the kinematic relation that (i) has in position dependence; (ii) symmetrical symmetric relation in position.As shown in Figure 4 (a), blue limit represents the kinematic relation that has dependence between limbs, and brown limit represents the symmetric relation between limbs.These two kinds of relations can clearly give expression to personage's example, as shown in Figure 4 (b).
Competitive relation has suppressed conflict in coupling when being activated simultaneously, mainly considers two aspects: two alternative target that (i) have same limbs type can not be activated simultaneously; (ii) lap of two alternative target should only be compared once.
4) limbs matching candidate figure is solved, find the coupling of best human limbs.
Adopt in the present embodiment compound cluster sampling algorithm to carry out iterative to limbs matching candidate figure, find the coupling of best human limbs, thereby complete the algorithm of personage's comparison.
Above-described embodiments of the present invention, do not form limiting the scope of the present invention.Any amendment of having done within spiritual principles of the present invention, be equal to and replace and improvement etc., within all should being included in claim protection domain of the present invention.
Claims (7)
1. the personage's comparison method based on gang form cluster sampling coupling, is characterized in that, comprises the following steps:
1) input several single pictures and many people of width picture, wherein single in several single pictures is same people;
2) from several single pictures, extract single many example set shuttering T, and from many people picture, extract alternative target S set;
3) build limbs matching candidate figure G by single many example set shuttering T and alternative target S set;
4) limbs matching candidate figure G is solved, find the coupling of best human limbs.
2. personage's comparison method according to claim 1, is characterized in that, described step 2) from several single pictures, extract single many example set shuttering T, its single many example set shuttering T is:
；
N=6, human limbs is divided into 6 parts, is respectively head, trunk, upper arm, forearm, thigh and shank; Wherein, g represents the concrete human limbs that each is detected, and represents a five-tuple for this rectangle frame with a rectangle frame
express, t represents the type of these limbs, and x and y represent the coordinate at this limbs center, and θ represents the anglec of rotation of these limbs, and s represents the relative scale of these limbs;
The concrete extracting method of above-mentioned single many example set shuttering T is:
21) single picture in the vertical direction is divided into 4 layers, in every one deck, scans detection respectively with corresponding human limbs detecting device, is detected accordingly mark, and concrete layered approach is as follows:
Ground floor: head
The second layer: trunk, upper arm, forearm
The 3rd layer: thigh
The 4th layer: shank
22) reject the limbs that are less than threshold X in testing result with the overlapping ratio of foreground mask, by remaining limbs composition template T.
3. personage's comparison method according to claim 1, it is characterized in that, described step 2) extracting method that extracts alternative target S set from many people picture is: scan many people picture with human limbs detecting device, obtain testing result, reject according to foreground mask the limbs that detect that are less than threshold X with prospect overlapping area again, by remaining composition alternative target S set.
4. according to the personage's comparison method described in claim 2 or 3, it is characterized in that, described threshold X is 75%.
5. according to the personage's comparison method described in claim 2 or 3, it is characterized in that, described foreground mask is to generate or manually mark generation by the method for background modeling.
6. personage's comparison method according to claim 1, it is characterized in that, the coupling that the summit of limbs matching candidate figure G in described step 3) is defined as corresponding limbs in single many example set shuttering T and target alternative S set is right, and the limit of limbs matching candidate figure G is defined as compatibility relation and the competitive relation on adjacent two summits;
Described compatibility relation is to encourage coupling to mutually activating in the process of coupling, and described compatibility relation is expressed as the close degree of two alternative target limb parts, mainly contains two aspects: the kinematic relation that (a) has in position dependence; (b) symmetrical symmetric relation in position;
Described competitive relation is to have suppressed conflict in coupling when being activated simultaneously, mainly contains two aspects: two alternative target that (i) have same limbs type can not be activated simultaneously; (ii) lap of two alternative target should only be compared once.
7. personage's comparison method according to claim 1, is characterized in that, described step 4) is to use compound cluster sampling algorithm to solve limbs matching candidate figure G, finds the coupling of best human limbs.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201410121866.8A CN103984916A (en) | 2014-03-28 | 2014-03-28 | Person comparison method based on combined template cluster sampling matching |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201410121866.8A CN103984916A (en) | 2014-03-28 | 2014-03-28 | Person comparison method based on combined template cluster sampling matching |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN103984916A true CN103984916A (en) | 2014-08-13 |
Family
ID=51276878
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201410121866.8A Pending CN103984916A (en) | 2014-03-28 | 2014-03-28 | Person comparison method based on combined template cluster sampling matching |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN103984916A (en) |
Cited By (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN106803056A (en) * | 2015-11-26 | 2017-06-06 | 华为技术有限公司 | The method of estimation and device of a kind of limbs relation |
| CN107451550A (en) * | 2016-03-15 | 2017-12-08 | 广东欧珀移动通信有限公司 | The method and Related product of unlocked by fingerprint |
Citations (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN103226694A (en) * | 2013-03-28 | 2013-07-31 | 赵福辉 | Real-time portrait acquiring, comparing and prewarning surveillance system and using method thereof |
| US8561193B1 (en) * | 2010-05-17 | 2013-10-15 | Symantec Corporation | Systems and methods for analyzing malware |
-
2014
- 2014-03-28 CN CN201410121866.8A patent/CN103984916A/en active Pending
Patent Citations (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US8561193B1 (en) * | 2010-05-17 | 2013-10-15 | Symantec Corporation | Systems and methods for analyzing malware |
| CN103226694A (en) * | 2013-03-28 | 2013-07-31 | 赵福辉 | Real-time portrait acquiring, comparing and prewarning surveillance system and using method thereof |
Non-Patent Citations (2)
| Title |
|---|
| LIANG LIN等: ""Layered graph matching by composite cluster sampling with collaborative and competitive interactions"", 《COMPUTER VISION AND PATTERN RECOGNITION》 * |
| YUANLU XU等: ""Human Re-identification by Matching Compositional Template with Cluster Sampling"", 《2013 IEEE INTERNATIONAL CONFERENCE ON COMPUTER VISION》 * |
Cited By (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN106803056A (en) * | 2015-11-26 | 2017-06-06 | 华为技术有限公司 | The method of estimation and device of a kind of limbs relation |
| CN107451550A (en) * | 2016-03-15 | 2017-12-08 | 广东欧珀移动通信有限公司 | The method and Related product of unlocked by fingerprint |
| CN107451550B (en) * | 2016-03-15 | 2020-12-04 | Oppo广东移动通信有限公司 | Fingerprint unlocking method and related product |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN105453153B (en) | Traffic lights detects | |
| CN103996191B (en) | A kind of detection method of black and white chessboard table images angle point based on Least-squares minimization | |
| CN103729649B (en) | A kind of image rotation angle detection method and device | |
| CN106778614B (en) | A kind of human body recognition method and device | |
| CN107705288A (en) | Hazardous gas spillage infrared video detection method under pseudo- target fast-moving strong interferers | |
| CN103902970B (en) | Automatic fingerprint Attitude estimation method and system | |
| CN104751146B (en) | A kind of indoor human body detection method based on 3D point cloud image | |
| CN104778710B (en) | A kind of morphological images edge detection method based on quantum theory | |
| CN103530599A (en) | Method and system for distinguishing real face and picture face | |
| CN106981081A (en) | A kind of degree of plainness for wall surface detection method based on extraction of depth information | |
| CN106651837B (en) | White glass plate surface chipping defect inspection method | |
| CN102930534A (en) | Method for automatically positioning acupuncture points on back of human body | |
| CN104850842B (en) | The man-machine interaction method of mobile terminal iris recognition | |
| CN103093451A (en) | Checkerboard intersection recognition algorithm | |
| US20180373927A1 (en) | Electronic device and gesture recognition method applied therein | |
| CN106097354A (en) | A kind of combining adaptive Gauss Face Detection and the hand images dividing method of region growing | |
| CN108171674A (en) | For the vision correcting method of visual angle projector image | |
| CN106097261A (en) | Image processing method and device | |
| CN105512656A (en) | Palm vein image collection method | |
| CN103984916A (en) | Person comparison method based on combined template cluster sampling matching | |
| CN107346408A (en) | Age recognition methods based on face feature | |
| CN109359581A (en) | A kind of face registration method based on intelligent glasses | |
| Li et al. | Hand gesture recognition by stereo camera using the thinning method | |
| CN103984927B (en) | Information processing method and electronic equipment | |
| CN105957067B (en) | A kind of color image edge detection method based on color difference |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| C06 | Publication | ||
| PB01 | Publication | ||
| C10 | Entry into substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| WD01 | Invention patent application deemed withdrawn after publication | ||
| WD01 | Invention patent application deemed withdrawn after publication |
Application publication date: 20140813 |