Implementation And Research On Key Issues Of Wrestling Videos Analysis And Retrieval

Fast and efficient tactics analysis is an important technical support to the competitive sports. Effective tactics analysis system is significant to improve the level of athletes. In this thesis, we focus on implementation of the wrestling video analysis and retrival system and research on key issues of video analysis. The achievement could be utilized for training the athletes and developing tactics in the wrestling competition. The primary work of the thesis includes as follows:The deep learning based wrestling action classification is studied. Based on the Caffe structure, and using the network trained from ImageNet classification as the initial network, we obtain a CNN model for key action recognition by fine-tuning. The experimental results confirm the efficiency of the framework.We design and implement the wrestling video analysis and retrieval system. It includes the wrestling video analysis system and the online wrestling match information retrieval system on B/S architecture. The platform has been utilized for tainging atheletes and supporting competetion in the national wrestling team.The wrestling video analysis system could be used to mark, collect and store the wrestling videos with human interfacing. And it could output a complete competition analysis report to present the key action and the score information. On one hand, the system could collect the technical data, on the other hand, we could make the effective and suitable training program with this data.The online wrestling match information retrieval system could access the video and technique report on line. The user with the access authority could logon and online browse the competition analysis reports saved on the server. And the corresponding video could be and study all of the collected sport video and the corresponding videos could be played back. By reviewing the competition videos related to specific atheletes, the coaches could make the more targetted tactics, which is helpful to improve the rate of winning.