Human Body Pose Intelligent Estimation Based on BlazePose

Intelligent physical fitness testing has always been the focus of researchers' attention, because of the large number of people participating in physical fitness testing and each physical fitness testing needs a lot of manpower and time, so the market has a strong demand for automated physical fitness testing equipment. In this paper, computer vision and deep learning methods are combined to extract key pose points from video information obtained by the camera using BlazePose. Then, filtering and feature extraction are carried out on these key points of attitude data and three machine learning methods are used to classify these data. The experimental results show that the action recognition rate obtained by ANN algorithm is 97.5%, and the accuracy of counting algorithm is 97.9%, which has great practical application value.