The research object of this paper is the sports game video. The athlete's motion state is clearly displayed in the video, and the target detection and tracking for the specified athlete is realized. A T-S fuzzy semantic interactive maneuvering target tracking algorithm based on two-dimensional fuzzy semantic reduction is proposed. Firstly, a knowledge system is composed of the weights of each model and the parameters of the previous ones, and the redundant models are eliminated by neighborhood fuzzy intensive reduction. At the same time, excessive fuzzy intensive reduction will consume a lot of computing power and lose effective information. An adaptive reduction judgment algorithm is proposed according to each moment and each model residual. Through residual monitoring and tracking, feature re-reduction and feature re-reduction can be adaptive. The experimental results show that the 2D fuzzy semantics can effectively reduce the features and improve the real-time performance of the algorithm. The test results show that the proposed method can effectively improve false alarm rate and tracking accuracy, and has higher robustness than traditional target detection and tracking in real complex competition venues.