Adaptive Recognition of Sports Image Features Based on Kalman Filtering Algorithm

In order to locate and identify the feature points of sports images quickly, an adaptive feature identification method of sports images based on Kalman filtering algorithm is proposed. The multi-target tracking model of sports images is constructed, and the linear constant-speed motion estimation of the target is established by using Kalman filter algorithm to predict. The score of motion correlation is measured by calculating the distance between the prediction center and the target detection center. In dynamic scenes, the perceptual classifier is used to make matching decisions, the target frame prediction is corrected by affine matrix, and the affine transformation between images is calculated by sparse optical flow, so as to realize adaptive recognition of sports image characteristics. The simulation results show that this method can effectively solve the multi-target tracking problem of sports moving images, and the video frame rate is very high, and the performance of inter-frame motion difference is effectively offset, which improves the ability of automatic feature positioning and correction of sports moving images.