Research on Sports Image Recognition and Tracking Based on Computer Vision Technology

Human behavior recognition technology is a sub-branch of computer vision technology, which can recognize and track human actions in videos through computers, so it is widely used in sports images. Aiming at the shortcomings of low recognition accuracy of current human body recognition technology, a hybrid algorithm (S-D) is proposed by combining semi-naive Bayes classification algorithm (SNBC) and dense trajectory algorithm (DT). The research results show that the number of training times for SNBC to achieve the target accuracy is 1762 less than that of the naive Bayes algorithm; the recognition accuracy of the S-D algorithm reaches 92%. The above results prove that the S-D algorithm can effectively improve the recognition accuracy of sports images.