Research on tracking and decomposing method of aerobics movement based on machine learning

Abstract

In order to overcome the low success rate of tracking and decomposing traditional aerobics movement tracking and decomposition method, this paper proposes a method of aerobics action tracking and decomposition based on machine learning. In this method, a multi-layer pyramid structure is built to segment the aerobics video image, and the key frame and the frame with the highest energy value are detected to obtain the binary image. This paper constructs a standardised tracking target model, uses convolution filter to detect multi-objective feature vectors in the image, uses binary classifier to build machine learning framework, and realises aerobics action tracking decomposition under the vector constraints of impact strength, spatial and frequency domain vectors. The experimental results show that this method can effectively deal with the interference from the external environment in partial sequence tracking, and compared with the traditional method, this method has higher success rate of tracking and decomposition and has reliability.