CN106991685B - Algorithm for tracking football in ball game video in panoramic mode - Google Patents
Algorithm for tracking football in ball game video in panoramic mode Download PDFInfo
- Publication number
- CN106991685B CN106991685B CN201610037570.7A CN201610037570A CN106991685B CN 106991685 B CN106991685 B CN 106991685B CN 201610037570 A CN201610037570 A CN 201610037570A CN 106991685 B CN106991685 B CN 106991685B
- Authority
- CN
- China
- Prior art keywords
- football
- tracking
- ball
- short
- track
- 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.)
- Active
Links
Images
Classifications
-
- 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
- G06T2207/10—Image acquisition modality
- G06T2207/10016—Video; Image sequence
-
- 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
- G06T2207/10—Image acquisition modality
- G06T2207/10024—Color image
-
- 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
- G06T2207/30241—Trajectory
Abstract
The invention discloses an algorithm for tracking a football in a ball game video in a panoramic mode, which comprises the following steps of 1) generating scene labels; 2) detecting candidate balls; 3) generating a short track; 4) rectification and filtering of short trajectories; 5) connecting the short tracks to the long tracks. The algorithm for tracking the football in the football match video in the panoramic mode extracts the corresponding motion characteristics of the football by utilizing some prior information (such as the position of a sports field of the football field, the position of a goal, the average height of players and the like) of the football field in the football motion and a plurality of main motion forms (ground pass, upward-hanging pass and player with ball) of the football, and adds the motion characteristics into the tracking of the football so as to improve the tracking accuracy.
Description
Technical Field
The invention relates to an algorithm for tracking a match video, in particular to an algorithm for tracking a football in a ball match video in a panoramic mode.
Background
When a common camera is adopted to record a football game without supervision, the problems of motion blur, ghost and ghost of a football which moves at a high speed are often caused due to the over-high speed of the football and the quality problem of the adopted camera lens; meanwhile, the situation that the ball is contacted with the players or is shielded by the players often occurs, and the area of the football is small in the panoramic mode, so that the football tracking effect of the general target tracking method in the ball game is poor. Aiming at the defects of the football tracking technology in the ball game video in the prior art, the invention provides an algorithm for tracking the football in the ball game video in the panoramic mode, so as to overcome the defects in the prior art.
Disclosure of Invention
The technical scheme adopted by the invention for solving the technical problems is to provide an algorithm for tracking a football in a ball game video in a panoramic mode, wherein the algorithm comprises the following steps: comprises the following 5 steps:
1) generating a scene label, and acquiring prior information of the scene by utilizing manual label or algorithm;
2) detecting candidate balls;
(1) establishing a background model for the panoramic video, and segmenting a foreground from the background model;
(2) filtering the foreground by using the size, shape and color information to obtain a candidate suspicious football area;
(3) predicting the position of the football in the current frame by using the result of tracking the football in the early stage, and if the position has a foreground, adding the position as a candidate suspicious football area;
(4) detecting a football in a suspicious football area by using a football model which is learned by an offline machine, and taking the area in which the football is detected as a candidate football area;
3) generating a short track;
(1) tracking the area by taking the currently detected candidate football as the area to be tracked, forming a track by a tracking result, and replacing the tracking result with a prediction result if the tracking result is not tracked; if the track is not tracked for a continuous period of time, ending the tracking of the track;
(2) detecting a human body in an expansion area of a currently detected and tracked ball area;
4) rectification and filtering of short trajectories; judging the movement of the track by using the prior information of the scene and the information detected by the human body and combining the movement rule of the football, and correcting and filtering the changed track according to the expected movement;
5) connecting the short tracks to long tracks; the short tracks are connected into long tracks by utilizing the time sequence and the spatial relationship to form more complete football tracking tracks.
The above algorithm for tracking a football in a video of a ball game in a panoramic mode, wherein: in the step of generating the scene marking, the prior information of the obtained scene comprises the steps of segmenting a football field from the video, extracting the goal position from the video and marking the heights of common players far away from the football field and near the football field in the video.
The above algorithm for tracking a football in a video of a ball game in a panoramic mode, wherein:
in the short trajectory rectification and filtering step, the trajectory is actively determined as follows:
if the central coordinate of the ball in the short track is approximate to a straight line, the ball is judged to be a ground pass;
if the central coordinate of the short track ball is similar to a parabola with a downward opening, the short track ball is judged to be an up-hanging pass ball;
if a human body exists near the center of the ball in the short track and the y coordinate is far lower than the height of the player at the position, the player is judged to take the ball.
The above algorithm for tracking a football in a video of a ball game in a panoramic mode, wherein:
in the short trajectory rectification and filtering step, where the trajectory is rectified and filtered according to the expected activity, the following are performed:
if the current ground pass period is judged according to the short track, correcting or directly rejecting the detection or tracking result which does not accord with the ground pass rule;
if the current hanging pass period is judged according to the short track, correcting or directly eliminating the detection or tracking result which does not accord with the hanging pass rule;
if the current time is judged to be the period of the player with the ball according to the short track, the detection or tracking result which does not accord with the player with the ball is corrected or directly eliminated.
Compared with the prior art, the invention has the following beneficial effects:
the method comprises the steps of extracting corresponding motion characteristics of a football by using some prior information (such as the position of a playing field of the football field, the position of a goal, the average height of players and the like) of the football field in the football and a plurality of main motion forms (ground pass, upward-hanging pass and player with ball) of the football, and adding the corresponding motion characteristics into the tracking of the football to improve the tracking accuracy.
Drawings
Fig. 1 is a schematic diagram illustrating the principle of an algorithm for tracking a soccer ball in a video of a ball game in a panoramic mode according to the present invention.
Detailed Description
The invention is further described below with reference to the figures and examples.
The invention provides an algorithm for tracking a football in a ball game video in a panoramic mode, which comprises the following specific schemes:
generating a scene annotation
And acquiring some prior information of the obtained scene by using a manual labeling or algorithm automatic learning method (for example, segmenting a football field from the video, extracting a goal position from the video, and labeling the heights of common players far away from the football field and near the football field in the video).
Detection of candidate balls
(1) Establishing a background model for the panoramic video, and segmenting a foreground from the background model;
(2) filtering the foreground by using information such as size, shape, color and the like to obtain some candidate suspicious football areas;
(3) predicting the position of the football in the current frame by using the result of tracking the football in the early stage, and if the position has a foreground, adding the position as a candidate suspicious football area;
(4) and (4) detecting the football in the suspicious football area by using the football model which is learned by the offline machine, and taking the area in which the football is detected out as a candidate football area.
Generation of short trajectories
(1) And (4) taking the currently detected candidate football as a region to be tracked, tracking the region (actually tracking and detecting to obtain a tracking result), and forming a track by the tracking result. If not, replacing the predicted result; if the track is not tracked for a continuous period of time, ending the tracking of the track;
(2) and detecting the human body in the expansion area of the currently detected and tracked ball area.
Short trajectory rectification and filtering
The prior information of the scene and the information of human body detection are utilized, the movement rule of the football is combined, the movement judgment is carried out on the track, and the track is corrected and filtered according to the expected movement.
Wherein, the track is judged by the following steps:
if the central coordinate of the ball in the short track is approximate to a straight line, the ball is judged to be a ground pass;
if the central coordinate of the short track ball is similar to a parabola with a downward opening, the short track ball is judged to be an up-hanging pass ball;
if a human body exists near the center of the ball in the short track and the y coordinate is far lower than the height of the player at the position, the player is judged to take the ball.
Wherein the trajectory is rectified and filtered according to the expected activity, as follows:
if the current ground pass period is judged according to the short track, correcting or directly rejecting the detection or tracking result (the central coordinate of the ball is approximate to a straight line) which does not accord with the ground pass rule;
if the current period is the period of hanging and passing the ball according to the short track, correcting or directly eliminating the detection or tracking result which does not conform to the rule of hanging and passing the ball (the track of the ball should be approximate to a parabola with a downward opening);
if the current time is judged to be the period of the player carrying the ball according to the short track, correcting or directly eliminating the detection or tracking result which does not accord with the player carrying the ball (the y coordinate of the center of the ball is far lower than the height of the player at the position);
connecting short tracks to long tracks
The short tracks are connected into long tracks by utilizing the time sequence and the spatial relationship to form more complete football tracking tracks.
Although the present invention has been described with respect to the preferred embodiments, it will be understood by those skilled in the art that various changes in form and details may be made therein without departing from the spirit and scope of the invention as defined by the appended claims.
Claims (3)
1. An algorithm for tracking a football in a video of a ball game in a panoramic mode, characterized in that: comprises the following 5 steps:
1) generating a scene label, and acquiring prior information of the scene by utilizing manual label or algorithm;
2) detecting candidate balls;
(1) establishing a background model for the panoramic video, and segmenting a foreground from the background model;
(2) filtering the foreground by using the size, shape and color information to obtain a candidate suspicious football area;
(3) predicting the position of the football in the current frame by using the result of tracking the football in the early stage, and if the position has a foreground, adding the position as a candidate suspicious football area;
(4) detecting a football in a suspicious football area by using a football model which is learned by an offline machine, and taking the area in which the football is detected as a candidate football area;
3) generating a short track;
(1) tracking the area by taking the currently detected candidate football as the area to be tracked, forming a track by a tracking result, and replacing the tracking result with a prediction result if the tracking result is not tracked; if the track is not tracked for a continuous period of time, ending the tracking of the track;
(2) detecting a human body in an expansion area of a currently detected and tracked ball area;
4) rectification and filtering of short trajectories; judging the movement of the track by using the prior information of the scene and the information detected by the human body and combining the movement rule of the football, and correcting and filtering the changed track according to the expected movement;
5) connecting the short tracks to long tracks; connecting the short tracks into long tracks by using the time sequence and the spatial relationship to form more complete football tracking tracks;
in the step of generating the scene marking, the prior information of the obtained scene comprises the steps of segmenting a football field from the video, extracting the goal position from the video and marking the heights of common players far away from the football field and near the football field in the video.
2. An algorithm for tracking a football in a video of a ball game in panoramic mode as claimed in claim 1, wherein: in the short trajectory rectification and filtering step, the trajectory is actively determined as follows:
if the central coordinate of the ball in the short track is approximate to a straight line, the ball is judged to be a ground pass;
if the central coordinate of the short track ball is similar to a parabola with a downward opening, the short track ball is judged to be an up-hanging pass ball;
if a human body exists near the center of the ball in the short track and the y coordinate is lower than the height of the player at the position, the player is judged to take the ball.
3. An algorithm for tracking a football in a video of a ball game in panoramic mode as claimed in claim 2, wherein:
in the short trajectory rectification and filtering step, where the trajectory is rectified and filtered according to the expected activity, the following are performed:
if the current ground pass period is judged according to the short track, correcting or directly rejecting the detection or tracking result which does not accord with the ground pass rule;
if the current hanging pass period is judged according to the short track, correcting or directly eliminating the detection or tracking result which does not accord with the hanging pass rule;
if the current time is judged to be the period of the player with the ball according to the short track, the detection or tracking result which does not accord with the player with the ball is corrected or directly eliminated.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201610037570.7A CN106991685B (en) | 2016-01-20 | 2016-01-20 | Algorithm for tracking football in ball game video in panoramic mode |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201610037570.7A CN106991685B (en) | 2016-01-20 | 2016-01-20 | Algorithm for tracking football in ball game video in panoramic mode |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN106991685A CN106991685A (en) | 2017-07-28 |
| CN106991685B true CN106991685B (en) | 2020-09-25 |
Family
ID=59413626
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201610037570.7A Active CN106991685B (en) | 2016-01-20 | 2016-01-20 | Algorithm for tracking football in ball game video in panoramic mode |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN106991685B (en) |
Families Citing this family (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN108090421B (en) * | 2017-11-30 | 2021-10-08 | 睿视智觉（深圳）算法技术有限公司 | Athlete athletic ability analysis method |
| CN108009504B (en) * | 2017-12-04 | 2020-06-26 | 深圳市赢世体育科技有限公司 | Moving sphere identification method and device and storage medium |
| US11610399B2 (en) | 2018-09-27 | 2023-03-21 | Intel Corporation | Highlight moment identification technology in volumetric content creation systems |
| CN110543856B (en) * | 2019-09-05 | 2022-04-22 | 新华智云科技有限公司 | Football shooting time identification method and device, storage medium and computer equipment |
| CN113781523B (en) * | 2021-09-13 | 2024-04-26 | 浙江大学 | Football detection tracking method and device, electronic equipment and storage medium |
Citations (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN101794451A (en) * | 2010-03-12 | 2010-08-04 | 上海交通大学 | Tracing method based on motion track |
| CN102142147A (en) * | 2010-01-29 | 2011-08-03 | 索尼公司 | Device and method for analyzing site content as well as device and method for detecting and tracking target |
| CN102819749A (en) * | 2012-07-23 | 2012-12-12 | 西安体育学院 | Automatic identification system and method for offside of football based on video analysis |
| CN104881882A (en) * | 2015-04-17 | 2015-09-02 | 广西科技大学 | Moving target tracking and detection method |
-
2016
- 2016-01-20 CN CN201610037570.7A patent/CN106991685B/en active Active
Patent Citations (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN102142147A (en) * | 2010-01-29 | 2011-08-03 | 索尼公司 | Device and method for analyzing site content as well as device and method for detecting and tracking target |
| CN101794451A (en) * | 2010-03-12 | 2010-08-04 | 上海交通大学 | Tracing method based on motion track |
| CN102819749A (en) * | 2012-07-23 | 2012-12-12 | 西安体育学院 | Automatic identification system and method for offside of football based on video analysis |
| CN104881882A (en) * | 2015-04-17 | 2015-09-02 | 广西科技大学 | Moving target tracking and detection method |
Also Published As
| Publication number | Publication date |
|---|---|
| CN106991685A (en) | 2017-07-28 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN106991359B (en) | Algorithm for tracking basketball in ball game video in panoramic mode | |
| CN106991685B (en) | Algorithm for tracking football in ball game video in panoramic mode | |
| CN106991357A (en) | The shooting of automatic identification Basketball Match and the algorithm scored based on panoramic video | |
| US8160302B2 (en) | Image processing apparatus and method for estimating orientation | |
| CN103413120B (en) | Tracking based on object globality and locality identification | |
| US8233721B2 (en) | Image processing apparatus and method | |
| CN106991358A (en) | The algorithm that automatic identification football match based on panoramic video is scored | |
| CN110929596A (en) | Shooting training system and method based on smart phone and artificial intelligence | |
| CN102819847B (en) | Based on the movement locus extracting method of PTZ dollying head | |
| Teachabarikiti et al. | Players tracking and ball detection for an automatic tennis video annotation | |
| JP6742221B2 (en) | Object tracking device and its program | |
| CN104506819A (en) | Multi-camera real-time linkage mutual feedback tracing system and method | |
| Beetz et al. | Visually tracking football games based on TV broadcasts | |
| Hsu et al. | Coachai: A project for microscopic badminton match data collection and tactical analysis | |
| Myint et al. | Tracking a table tennis ball for umpiring purposes | |
| CN101715070A (en) | Method for automatically updating background in specifically monitored video | |
| CN105187801B (en) | System and method for generating abstract video | |
| Zita et al. | Tracking fast moving objects by segmentation network | |
| Kamiyama et al. | Improvement of badminton-player tracking applying image pixel compensation | |
| Christmas et al. | A system for the automatic annotation of tennis matches | |
| JP2020107071A (en) | Object tracking device and program thereof | |
| CN106326910A (en) | Video shot classified labeling method and video shot classified labeling device | |
| CN110717931B (en) | System and method for detecting height of hitting point of badminton serving | |
| Zhu et al. | hTLD: A Human-in-the-loop Target Detection and Tracking Method for UAV | |
| Kumar et al. | 3d estimation and visualization of motion in a multicamera network for sports |
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 | ||
| GR01 | Patent grant | ||
| GR01 | Patent grant | ||
| TR01 | Transfer of patent right | ||
| TR01 | Transfer of patent right |
Effective date of registration: 20220324 Address after: 200080 room 3001, building A2, No. 121, Zhongshan North 1st Road, Hongkou District, Shanghai Patentee after: Hupu (Shanghai) Information Technology Co.,Ltd. Address before: 200080 room 7179, ground floor, No. 203, Xinjian Road, Hongkou District, Shanghai Patentee before: SHANGHAI HUITI NETWORK TECHNOLOGY CO.,LTD. |