Human body attitude estimation method based on cascade error correction mechanism

Abstract

The invention discloses a human body attitude tracking method based on a cascade error correction mechanism and mainly solves a problem of inaccurate wrist positioning during human body attitude estimation in the prior art. The method comprises steps that 1), a bidirectional tree structure model is utilized to position each body articulation point except a wrist; 2), an optical flow and particle filtering are utilized to preliminarily predict the position of the wrist; 3), whether the preliminary detection result is reliable is determined through optical flow response, if not, a bidirectional graph structure model is utilized to position the position of the wrist; 4), a skin color model is utilized to determine whether the positioning result of the bidirectional graph structure model is reliable; and 5), if the positioning result of the bidirectional tree structure model is inaccurate, and the previous wrist position is re-utilized to estimate the present wrist position. The method is advantaged in that the wrist position can be more accurately positioned, the better human body attitude estimation effect can be acquired, and the method can be applied to human body attitude identification in occasions of video monitoring, man-machine interaction, digital entertainment, medical imaging and motion scenes.

Description

technical field [0001] The invention belongs to the technical field of image recognition, in particular to a method for estimating human body posture, which can be used to recognize human body posture in video surveillance, human-computer interaction, digital entertainment, medical imaging and sports scenes. Background technique [0002] Human body pose estimation refers to the process of detecting the position of each part of the human body in a 2D or 3D static image or video, and estimating the human body pose according to the connection relationship of each part of the human body. Human pose estimation is an important problem in the field of computer vision research, and it has broad application prospects in many practical applications, such as video surveillance, human-computer interaction, digital entertainment, medical imaging and sports scenes. But at the same time, accurate estimation of human pose is also a very challenging problem. In real life, changes in human b...

AI Technical Summary

Problems solved by technology

[0006] The purpose of the present invention is to solve the problem of inaccurate wrist positioning in human body posture estimation, and propose a human body posture estimation method based on a cascaded error correction mechanism by using the unique time information of video image sequences and an adaptive skin color model, so as to improve The correct detection rate of the wrist, to obtain a more accurate human body pose estimation effect

Images