Behavior Recognition Algorithm Based on Key Points of Human Bones

Abstract

Human behavior recognition method based on video image easily influenced by background factors with lead to recognition accuracy is not high, and in view of the traditional Shuangliu network needs computing optical flow chart in advance and consume large amounts of time and need a lot of space to store an image of a light flow, an improved fusion skeleton information and image information of human behavior recognition algorithm. The experimental results show that the algorithm can effectively improve the accuracy of behavior recognition in video, and improve the identification ability of time-dependent behaviors and approximate behaviors.