Research On Interactive Behavior Analysis For Soccer Video

Sports videos are one kind of the popular videos which contain rich topics of intelligent information processing in the fields of video content analysis. Since soccer video is a typical form of sports videos, analyzing and researching of which have important application value, especially in aspect of extracting high-level semantic information from the videos. Current research mainly focus on event retrieval, scene classification, moving target detection, etc., while the work about team evaluation, scene semantic analysis, auxiliary judgments, tactical analysis, etc. is not perfect, in addition, the event development thread or the event initiation process are always neglected. In order to achieve better performance on intelligent analysis of soccer video, the theories and methods of digital image processing, pattern classification, and computer vision were introduced in this dissertation to analyze the interaction behavior in soccer video. And furthermore, the key problems of soccer video semantic understanding system were studied from several parts, namely scene analysis, target detection, target recognition, target classification, target tracking, trajectory segmentation and strategy analysis, which made a useful exploration in understanding and analyzing the competition from a common scene.Firstly, an algorithm of player classification based on digraph was proposed according to the player segmentation. In the algorithm, the player images were extracted after scene analysis, and the player images then were divided into blocks adopting rectangular partition strategy after the color space quantification of HSV color model. According to the distance matrix calculated by the color feature of histogram statistics in each block, a digraph was generated. Then the player classification was implemented by classifying the vertexes of digraph.Secondly, a semantic judging method of team offensive goal was proposed based on the degree of defense. Considering the features of players’position and number of players, the number was calculated in the domain limits and the average position was calculated referring to players’ position. Then the degree of defense for each team was obtained after setting impact factor. Then the offense direction for both sides was discriminated according to the degree of defense.Finally, a method of offensive strategy analysis based on real target trajectory was proposed in order to analyze the offensive strategy exist in the soccer video according to the video features analysis. After extracting the real trajectory, a trajectory segmentation method based on the angle difference was put forward to analyze the interactions between players and the ball. And furthermore, eight typical attack modes were defined via analyzing the relationship between status and position of the ball. Meanwhile, the mode recognition method was propose in the end.Experiment results show that the proposed algorithms can effectively solve the problem of each stage in the video analysis and good performance is achieved from analyzing the interactions existing in the video. In addition, analysis of the offensive strategy is realized from the perspective of the soccer field. And good achievements are obtained in both the. application of computer vision research and the scientific research for soccer sports.