Research on Human Skeleton extraction algorithm based on Machine Learning

Abstract

Analyzing the human skeleton of the target and extracting the motion information is of great significance to the establishment of human model and the research and analysis of human motion. In this paper, an improved method for establishing the star skeleton model of the human body is proposed. The centroid extracted from the traditional star skeleton model moves up and down according to the proportion of the human torso to obtain the parameters of the central point of the shoulder and buttocks. And the combination of face-to-face mannequin and side-looking mannequin to build a more detailed mannequin. The experimental results show that this method improves the accuracy of locating the endpoints of human limbs and can better obtain the human skeleton model.