Algorithm for tracking player in match video

Abstract

The invention discloses an algorithm for tracking a player in a match video. The algorithm comprises four steps of initializing, image preprocessing, detecting a human body by a passenger detector, and fusing area information and local characteristics. The algorithm for tracking the player in the match video has advantages of combining advantages of area information and local characteristics, realizing high suitability for the condition of relatively severe shielding, effectively preventing a drift problem, effectively settling a problem of player tracking in a match, and overcoming a defect of relatively low motion tracking effect in the match according to an object tracking method in prior art.

Description

technical field

[0001] The invention relates to a tracking algorithm of a football game video, in particular to an algorithm for tracking athletes in a football game video. Background technique

[0002] When an ordinary camera is used for unsupervised recording of a ball game, due to the high speed of the players, it often leads to motion blur, ghosting and afterimage problems for high-speed athletes; at the same time, due to the large range of motion and frequent contact or In the case of occlusion, this makes the general object tracking method less effective for tracking motion in the ball game. The algorithm for tracking players in the ball game video proposed by the present invention has the advantages of integrating regional information and local features, can be adapted to serious occlusion situations, and effectively prevents the problem of drifting, and can effectively solve the problem of player tracking in a ball game. Contents of the invention

[0003] The tech...

Examples

Embodiment Construction

[0025] The present invention will be further described below in conjunction with the accompanying drawings and embodiments.

[0026] A kind of algorithm that the present invention provides is tracked to the athlete in the ball game video, and specific scheme comprises:

[0027] 1. Initialization.

[0028] Use manual annotation or algorithm automatic learning method to obtain some prior information of the scene (such as extracting the position of the court from the video, the height of ordinary players around the court).

[0029] 2. Image preprocessing.

[0030] (1) Establish a background model for the stadium in the video, and segment the foreground from it, and extract local feature points (such as FAST feature points) on the foreground;

[0031] (2) Extract the edge from the foreground area (such as using SOBEL edge extraction operator).

[0032] (3) For the feature points extracted from the previous frame, the position of the current frame is obtained by optical flow tra...

technical field

[0001] The invention relates to a tracking algorithm of a football game video, in particular to an algorithm for tracking athletes in a football game video. Background technique

[0002] When an ordinary camera is used for unsupervised recording of a ball game, due to the high speed of the players, it often leads to motion blur, ghosting and afterimage problems for high-speed athletes; at the same time, due to the large range of motion and frequent contact or In the case of occlusion, this makes the general object tracking method less effective for tracking motion in the ball game. The algorithm for tracking players in the ball game video proposed by the present invention has the advantages of integrating regional information and local features, can be adapted to serious occlusion situations, and effectively prevents the problem of drifting, and can effectively solve the problem of player tracking in a ball game. Contents of the invention

[0003] The tech...

Examples

Embodiment Construction

[0025] The present invention will be further described below in conjunction with the accompanying drawings and embodiments.

[0026] A kind of algorithm that the present invention provides is tracked to the athlete in the ball game video, and specific scheme comprises:

[0027] 1. Initialization.

[0028] Use manual annotation or algorithm automatic learning method to obtain some prior information of the scene (such as extracting the position of the court from the video, the height of ordinary players around the court).

[0029] 2. Image preprocessing.

[0030] (1) Establish a background model for the stadium in the video, and segment the foreground from it, and extract local feature points (such as FAST feature points) on the foreground;

[0031] (2) Extract the edge from the foreground area (such as using SOBEL edge extraction operator).

[0032] (3) For the feature points extracted from the previous frame, the position of the current frame is obtained by optical flow tra...

Images