Three-dimensional motion tracking method of upper half part of human body based on monocular video

Abstract

The invention discloses a three-dimensional motion tracking method of an upper half part of a human body based on a monocular video. The method comprises calculating via sift-matching characteristic points according to chest shape invariance to obtain a coordinate on an image, variable factors, a rotating Euler angle, etc. of chest joints, and obtaining rotating postures of left and right upper arms and forearms via an in-depth traverse method according to the skeleton model of the upper half part of human body. The method establishes an equation according to inverse kinematics and the coordinates and post-rotation coordinates of the matching characteristic points in a local coordinate system, and solves the equation to thus obtain the rotating Euler angle. The inventive method is simple, can accurately track the posture of the upper half part of a human body, and can be widely applied in the fields of human-machine interaction, interactive entertainment, intelligent monitoring, medical diagnosis, etc.

Description

【Technical field】 [0001] The invention relates to the fields of computer vision and video processing, in particular to a method for recovering human body three-dimensional motion posture based on monocular video. 【Background technique】 [0002] Video-based three-dimensional human body motion tracking has great application value in fields such as human-computer interaction, interactive entertainment, intelligent monitoring, and medical diagnosis. Different from binocular or multi-eye video human motion tracking, monocular video tracking only needs one camera to capture human body motion and does not require camera calibration, so it is easy to use and widely used. [0003] At present, the commonly used 3D human body motion tracking under monocular video mainly adopts learning-based methods. This method has the following disadvantages: 1) The use of statistical learning requires a large number of training samples, and the restored human body pose has a large uncertainty. 2) ...

AI Technical Summary

Problems solved by technology

This method has the following disadvantages: 1) The use of statistical learning requires a large number of training samples, and the restored human body pose has a large uncertainty

2) The effect of statistical learning tracking depends entirely on the training samples. Due to the uncertainty of human motion, it is difficult for the algorithm trained by a certain sample set to apply to all human motions, and the algorithm cannot be generalized

Images