CN113963375A - Multi-feature matching multi-target tracking method for fast skating athletes based on regions - Google Patents
Multi-feature matching multi-target tracking method for fast skating athletes based on regions Download PDFInfo
- Publication number
- CN113963375A CN113963375A CN202111220338.4A CN202111220338A CN113963375A CN 113963375 A CN113963375 A CN 113963375A CN 202111220338 A CN202111220338 A CN 202111220338A CN 113963375 A CN113963375 A CN 113963375A
- Authority
- CN
- China
- Prior art keywords
- matching
- speed skating
- speed
- tracking
- feature
- 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 29
- 238000001514 detection method Methods 0.000 claims abstract description 11
- 238000013135 deep learning Methods 0.000 abstract description 3
- 230000011218 segmentation Effects 0.000 abstract description 2
- 230000008569 process Effects 0.000 description 4
- 238000005259 measurement Methods 0.000 description 2
- 239000013598 vector Substances 0.000 description 2
- 101000642315 Homo sapiens Spermatogenesis-associated protein 17 Proteins 0.000 description 1
- 241001465754 Metazoa Species 0.000 description 1
- 102100036408 Spermatogenesis-associated protein 17 Human genes 0.000 description 1
- 238000013528 artificial neural network Methods 0.000 description 1
- 230000006399 behavior Effects 0.000 description 1
- 238000004364 calculation method Methods 0.000 description 1
- 230000008859 change Effects 0.000 description 1
- 238000010586 diagram Methods 0.000 description 1
- 230000003993 interaction Effects 0.000 description 1
- 238000000691 measurement method Methods 0.000 description 1
- 238000012986 modification Methods 0.000 description 1
- 230000004048 modification Effects 0.000 description 1
- 238000012544 monitoring process Methods 0.000 description 1
- 238000011160 research Methods 0.000 description 1
Images
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/22—Matching criteria, e.g. proximity measures
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T7/00—Image analysis
- G06T7/10—Segmentation; Edge detection
- G06T7/11—Region-based segmentation
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T7/00—Image analysis
- G06T7/20—Analysis of motion
- G06T7/246—Analysis of motion using feature-based methods, e.g. the tracking of corners or segments
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T2207/00—Indexing scheme for image analysis or image enhancement
- G06T2207/30—Subject of image; Context of image processing
- G06T2207/30196—Human being; Person
Landscapes
- Engineering & Computer Science (AREA)
- Theoretical Computer Science (AREA)
- Physics & Mathematics (AREA)
- Computer Vision & Pattern Recognition (AREA)
- General Physics & Mathematics (AREA)
- Data Mining & Analysis (AREA)
- Artificial Intelligence (AREA)
- Bioinformatics & Cheminformatics (AREA)
- Bioinformatics & Computational Biology (AREA)
- Evolutionary Biology (AREA)
- Evolutionary Computation (AREA)
- Life Sciences & Earth Sciences (AREA)
- General Engineering & Computer Science (AREA)
- Multimedia (AREA)
- Image Analysis (AREA)
Abstract
The invention combines deep learning and computer vision algorithm, and particularly discloses a multi-target tracking method for a multi-feature matching speed skating player based on a region, which comprises the following steps: s1, dividing the current area according to the speed skating site; s2, detecting the information of the fast skaters in the picture according to a target detection algorithm; s3, performing zone division on the speed skating player information detected in s2 according to the divided zones in s 1; s4, matching the speed skiers in each area by using the multi-feature information; and s5, matching the areas according to the tracking result of the speed skating players in each area to form a final tracking result. The method of the invention aims at the characteristic information of the speed skating player such as the match duration, the speed skating field and the like, and carries out the multi-target tracking of the speed skating player more robustly through the modes of region segmentation, multi-characteristic matching and the like.
Description
Technical Field
The invention combines deep learning and a computer vision algorithm, and particularly discloses a multi-target tracking method for a multi-feature matching speed skating player based on a region.
Background
The multi-target tracking is a key technology in the field of computer vision, and is widely applied in directions of automatic driving, intelligent monitoring, behavior recognition and the like. However, due to the complexity of the multi-target tracking task, the multi-target tracking task faces more challenges than the target detection and single-target tracking tasks, such as target overlapping, appearance sharp change, appearance similarity and the like. How to solve the problems more effectively has great significance for the application of multi-target tracking technology, so that in the past decades, people propose a wide range of solutions.
Multi Object Tracking (MOT) refers to detecting a plurality of objects such as pedestrians, automobiles, and animals in a video and giving IDs to perform trajectory Tracking without knowing the number of objects in advance. The main task is to locate a plurality of interested targets in a given video at the same time, and different targets have different IDs so as to realize subsequent track prediction, accurate search and other work. Compared with other computer vision tasks, the multi-target tracking task mainly has the following research difficulties: 1) target detection is not accurate enough; 2) frequent target occlusion; 3) the target number is uncertain; 4) similar appearance, interaction between multiple targets.
Nowadays, the multi-target tracking methods with the highest attention in the industry are mainly SORT and deep SORT. The two methods realize multi-target tracking through target detection, Hungary matching and Kalman prediction and updating. However, when the application is landed, various numerical values of the tracking target in a real scene, including the moving distance and moving speed of the tracking target per second, need to be obtained.
In the invention, a multi-feature matching multi-target tracking method for a speed skating player based on a region is provided. The tracking matching for skaters is quite different from the matching methods of existing data sets. In data sets for pedestrians such as the conventional data sets MOT17 and MOT19, most pedestrians perform linear motion, and the tracking time is relatively short. The fast-skating players have long competition time and move according to the field, so the sliding track is relatively fixed, and the existing algorithm can not be used for good matching. Based on the problems, the whole field is tracked regionally, so that the motion states of the speed skating players in each region are relatively consistent; meanwhile, the process of the previous frame and the next frame is better matched in a multi-feature mode. The track of the speed skating player can be better tracked through the multi-zone and multi-characteristic mode.
According to the invention, a multi-target tracking scheme is established by combining deep learning and computer vision methods, and a multi-target tracking method based on multi-region and multi-feature matching is carried out under a single camera, so that the method can be closer to the motion features of the fast-skating players, and the fast-skating players can be better tracked.
Disclosure of Invention
The invention aims to provide a multi-feature matching multi-target tracking method for a speed skating player based on a region, which adopts the following scheme:
a multi-target tracking method for a multi-feature matching speed skating player based on a region comprises the following steps:
s1, dividing the current area according to the speed skating site;
s2, detecting the information of the fast skaters in the picture according to a target detection algorithm;
s3, performing zone division on the speed skating player information detected in s2 according to the divided zones in s 1;
s4, matching the speed skiers in each area by using the multi-feature information;
and s5, matching the areas according to the tracking result of the speed skating players in each area to form a final tracking result.
Further, in the step s1, the current area is divided according to the characteristics of the speed skating field, so that the movement characteristics in each area are relatively fixed;
further, in the step s2, feature information such as the current position and posture of the fast-skating player in the picture is detected according to a target detection algorithm;
further, the following characteristics of the skater are detected:
s21, skater's location and bounding box information;
s22, characteristic dimension vector of speed skating player, for using measurement method to match ID of front and back frames;
s23 posture characteristics of the skater, comprising 17 joints representing the position information of the joints of the skater.
Further, in the above step s3, the speed skating player information detected in s2 is divided into zones according to the zones in s1, so that different tracking models can be applied to each zone;
further, in step s4, according to the information of the skater in each area, the information of the multiple characteristics is used for carrying out association matching on the frame before and after the skater;
further, in step s5, according to the tracking result of the skater in each area, matching between areas to form a final tracking result;
further, the matching process between the regions is as follows:
s51, obtaining the tracking result of the skater in the common part between the areas;
s52, obtaining the matching relation between the IDs in the front and rear regions in the time period through voting, matching the regions by using the tracking result of the common part to obtain the final matching result, and finally forming a complete tracking track.
The invention has the following advantages:
the method uses a deep neural network under computer vision, tracks multiple targets in real time through a detection and tracking mode, and simultaneously uses a multi-region method to divide a region of a speed skating field, so that the movement characteristics of speed skating players in the region accord with certain linear or curve characteristics. The method corrects the existing multi-target tracking method according to the motion characteristics of the existing fast skating players, so that the multi-target tracking task of the fast skating players can be completed more quickly and better.
Drawings
FIG. 1 is a flow chart of an implementation of a region-based multi-feature matching skater multi-target tracking method;
FIG. 2 is a schematic diagram of zone segmentation for a zone-based multi-feature matching skater multi-target tracking method;
detailed description of the invention
The invention will be described in further detail with reference to the accompanying figure 1 and the following detailed description:
referring to fig. 1, a multi-target tracking method for a multi-feature matching speed skating player based on zones includes the following steps:
s1, dividing the current area according to the speed skating site, wherein the divided areas are shown in the attached figure 2;
in order to improve the detection accuracy and obtain a better multi-target tracking result, the sports field is divided into 8 areas according to the motion characteristics of different speed gliders in the field, and pure linear or curve tasks are carried out among the areas.
s2, detecting the information of the fast skaters in the picture according to a target detection algorithm;
in order to better apply various characteristics to track and match the skater, various information of the skater in the picture is utilized in the target detection stage.
s21, the position of the skater and bounding box information, primarily indicating the position of the skater in the picture, and the area in which it is located;
s22, characteristic information of the speed skating players, wherein the motion characteristic information is a group of pedestrian re-identification vectors with the length of N and is used for carrying out characteristic measurement.
s23 posture characteristics of the skater, comprising 17 joints representing the position information of the joints of the skater.
s3, performing zone division on the detected speed skating player information according to the zone in s1, wherein the specific division principle is as follows:
s31, marking the position of the regional characteristic point in the field in the picture, and calculating the regional division used by 4 curves for motion according to the corresponding characteristic point.
The area division process corresponding to the s32 and 4 curves is as follows;
the calculation process is as follows:
1. the real field corresponding to the plane image is an ellipse, and is divided into two parts according to the major axis curve, wherein the left part is marked as 0, and the right part is marked as 1;
2. the three curves are equally taken in parallel with the short axis, the upper part of the first curve is marked as 0, and the lower part of the first curve is marked as 1;
3. the continuation above the second curve is marked as 0 and the continuation below is marked as 1;
4. the continuation above the third curve is marked as 0 and the continuation below is marked as 1;
5. the corresponding eight blocks of regions get the corresponding serial numbers of 0000, 0100, 0110, 0111, 1000, 1100, 1110, 1111.
s33, calculating the area position of the speed skating player according to the position information of the speed skating player, and applying different area matching principles.
s4, matching the speed skiers in each area by using the multi-feature information;
and performing comprehensive characteristic matching measurement by using the region position, the pedestrian re-identification characteristic and the attitude joint point characteristic according to the information of the fast-skating players in each region, and performing association matching on the front frame and the rear frame of the fast-skating players.
And s5, matching the areas according to the tracking result of the speed skating players in each area to form a final tracking result.
Acquiring the tracking result of the common part of the fast skaters between the areas, acquiring the matching relation between the IDs in the front area and the back area in a voting mode in a time period, and matching the areas by using the tracking result of the common part to obtain the final matching result, thereby finally forming a complete tracking track.
It should be understood that the above description is only an overall implementation flow of the present invention, and the present invention is not limited to the above implementation flow, and it should be noted that all equivalent and obvious modifications made by those skilled in the art under the teaching of the present specification fall within the spirit scope of the present specification, and should be protected by the present invention.
Claims (3)
1. A multi-target tracking method for a multi-feature matching speed skating player based on a region is characterized by comprising the following steps:
s1, dividing the current area according to the speed skating site;
s2, detecting the information of the fast skaters in the picture according to a target detection algorithm;
s3, performing zone division on the speed skating player information detected in s2 according to the divided zones in s 1;
s4, matching the speed skiers in each area by using the multi-feature information;
and s5, matching the areas according to the tracking result of the speed skating players in each area to form a final tracking result.
2. The multi-target tracking method for the speed skating players based on the zone of claim 1, wherein in the steps s1 and s3, the sports ground is divided into zones, and a multi-feature tracking strategy is applied to different zones, so that the tracking performance of the speed skating players is improved.
3. The method as claimed in claim 1, wherein the step s4 is performed by performing a comprehensive metric matching between frames according to the multi-feature matching technique, the region location information, the pedestrian re-identification feature and the posture feature.
Priority Applications (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN202111220338.4A CN113963375A (en) | 2021-10-20 | 2021-10-20 | Multi-feature matching multi-target tracking method for fast skating athletes based on regions |
Applications Claiming Priority (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN202111220338.4A CN113963375A (en) | 2021-10-20 | 2021-10-20 | Multi-feature matching multi-target tracking method for fast skating athletes based on regions |
Publications (1)
Publication Number | Publication Date |
---|---|
CN113963375A true CN113963375A (en) | 2022-01-21 |
Family
ID=79465643
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
CN202111220338.4A Pending CN113963375A (en) | 2021-10-20 | 2021-10-20 | Multi-feature matching multi-target tracking method for fast skating athletes based on regions |
Country Status (1)
Country | Link |
---|---|
CN (1) | CN113963375A (en) |
Citations (8)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN1738426A (en) * | 2005-09-09 | 2006-02-22 | 南京大学 | Video motion goal division and track method |
US20110115920A1 (en) * | 2009-11-18 | 2011-05-19 | Industrial Technology Research Institute | Multi-state target tracking mehtod and system |
CN109410245A (en) * | 2018-09-13 | 2019-03-01 | 北京米文动力科技有限公司 | A kind of video target tracking method and equipment |
CN109448025A (en) * | 2018-11-09 | 2019-03-08 | 国家体育总局体育科学研究所 | Short-track speeding skating sportsman's automatically tracks and track modeling method in video |
CN110490901A (en) * | 2019-07-15 | 2019-11-22 | 武汉大学 | The pedestrian detection tracking of anti-attitudes vibration |
CN111639570A (en) * | 2020-05-20 | 2020-09-08 | 华中科技大学 | Online multi-target tracking method based on motion model and single-target clue |
CN111709974A (en) * | 2020-06-22 | 2020-09-25 | 苏宁云计算有限公司 | Human body tracking method and device based on RGB-D image |
CN111882580A (en) * | 2020-07-17 | 2020-11-03 | 元神科技(杭州)有限公司 | Video multi-target tracking method and system |
-
2021
- 2021-10-20 CN CN202111220338.4A patent/CN113963375A/en active Pending
Patent Citations (8)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN1738426A (en) * | 2005-09-09 | 2006-02-22 | 南京大学 | Video motion goal division and track method |
US20110115920A1 (en) * | 2009-11-18 | 2011-05-19 | Industrial Technology Research Institute | Multi-state target tracking mehtod and system |
CN109410245A (en) * | 2018-09-13 | 2019-03-01 | 北京米文动力科技有限公司 | A kind of video target tracking method and equipment |
CN109448025A (en) * | 2018-11-09 | 2019-03-08 | 国家体育总局体育科学研究所 | Short-track speeding skating sportsman's automatically tracks and track modeling method in video |
CN110490901A (en) * | 2019-07-15 | 2019-11-22 | 武汉大学 | The pedestrian detection tracking of anti-attitudes vibration |
CN111639570A (en) * | 2020-05-20 | 2020-09-08 | 华中科技大学 | Online multi-target tracking method based on motion model and single-target clue |
CN111709974A (en) * | 2020-06-22 | 2020-09-25 | 苏宁云计算有限公司 | Human body tracking method and device based on RGB-D image |
CN111882580A (en) * | 2020-07-17 | 2020-11-03 | 元神科技(杭州)有限公司 | Video multi-target tracking method and system |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
CN106022220B (en) | Method for tracking multiple faces of participating athletes in sports video | |
Breitenstein et al. | Online multiperson tracking-by-detection from a single, uncalibrated camera | |
McNally et al. | Golfdb: A video database for golf swing sequencing | |
Wen et al. | DETRAC: A new benchmark and protocol for multi-object tracking | |
CN109191497A (en) | A kind of real-time online multi-object tracking method based on much information fusion | |
Zhang et al. | A sport athlete object tracking based on deep sort and yolo V4 in case of camera movement | |
CN107315994B (en) | Clustering method based on Spectral Clustering space trajectory | |
CN103886619A (en) | Multi-scale superpixel-fused target tracking method | |
Teachabarikiti et al. | Players tracking and ball detection for an automatic tennis video annotation | |
CN115240130A (en) | Pedestrian multi-target tracking method and device and computer readable storage medium | |
CN112270226A (en) | Pedestrian trajectory prediction method based on multi-feature extraction and multi-attention mechanism | |
CN106296743A (en) | A kind of adaptive motion method for tracking target and unmanned plane follow the tracks of system | |
Ivasic-Kos et al. | Building a labeled dataset for recognition of handball actions using mask R-CNN and STIPS | |
CN113256690B (en) | Pedestrian multi-target tracking method based on video monitoring | |
Sha et al. | Understanding and analyzing a large collection of archived swimming videos | |
CN106327528A (en) | Moving object tracking method and operation method of unmanned aerial vehicle | |
Naik et al. | DeepPlayer-track: player and referee tracking with jersey color recognition in soccer | |
Zhang et al. | Real-time golf ball detection and tracking based on convolutional neural networks | |
De Langis et al. | Realtime multi-diver tracking and re-identification for underwater human-robot collaboration | |
CN115731266A (en) | Cross-camera multi-target tracking method, device and equipment and readable storage medium | |
CN113947108A (en) | Player tracking detection method based on YOLO V5 | |
Ludwig et al. | Robust estimation of flight parameters for ski jumpers | |
Hsu et al. | Coachai: A project for microscopic badminton match data collection and tactical analysis | |
CN114283355A (en) | Multi-target endangered animal tracking method based on small sample learning | |
CN110659544A (en) | Milk cow behavior identification method based on nonparametric spatiotemporal context trajectory model |
Legal Events
Date | Code | Title | Description |
---|---|---|---|
PB01 | Publication | ||
PB01 | Publication | ||
SE01 | Entry into force of request for substantive examination | ||
SE01 | Entry into force of request for substantive examination |