In this paper, we study the operation mechanism of the Openpose model under artificial intelligence technology, reconstruct the spatial coordinate system of the skeleton, and do the standardization of the skeleton coordinate system. Analyze the problem of limb angle similarity, sort out the algorithm process for extracting differential posture features, and extract human limb angle features through center-of-gravity regression calculation. A model for the change of non-heritage inheritance methods in sports is constructed using motion-captured interactive images of non-heritage sports dance. There are five innovative ways to use non-heritage protection strategies. The Openpose model was used to capture the kinematic features of sports gestures to analyze the non-genetic inheritance of sports in A place for empirical research. In the study of athletes’ front leg characteristics, the mean value of level 1 athletes was 44.68±21.93 higher than that of level 2 athletes, and the p-value was 0.049<0.05, which was significant. In the empirical analysis of sports NRLs, the comprehensive score of the protection effect of 14 sports NRLs in place A was 3.293, which was better overall. Among the protection effects of different dimensions, the survival conditions dimension has the best protection effect, with a mean value of 4.436. This aspect of survival conditions is handled with the best effect in place A. The government is not as good as the other two dimensions in terms of protection and needs to increase its protection efforts.