An Efficient Method of Human Motion Analysis on Dual-Camera Stereo Vision System

Abstract

An efficient method to analyze human motion is proposed based on object segmentation and stereo vision. First, the object region segmentation and binary image skeletonization are accomplished. Then the key points on the object outline are located, and the matching of key points is accomplished by the normalized color cross-correlation as the measure of similarity. The parallax values of key points are estimated from the matching results, which can be used to analyze human motion. The results indicate effectiveness and efficiency of the method for human motion analysis.