RECOGNITION METHOD OF DIVERSIFIED TEACHING MODE OF COLLEGE PHYSICAL TRAINING BASED ON DATA MINING

In the current sports training process, trainers use the less difficult video replay and analysis management method to explain the action essentials for athletes, which is not intuitive and scientific enough to meet the needs of trainers for sports effect evaluation. In view of this problem, this paper studies the sports video analysis and recognition process in the process of sports training, and uses the adaptive threshold moving target separation algorithm based on particle filter prediction to achieve the adaptive separation of moving targets. Through the particle filter technology to track the athletes' motion, shape the motion model, and predict the position of different important joint points in the follow-up motion video frame according to the motion model to complete the follow-up motion video frame tracking. The method of conditional random field is used to realize the action recognition in sports training video. Experimental results show that this method has high action recognition rate and low false separation rate.