Study on Stereo Vision-based Parameterized Virtual Hand Modeling

Abstract

Virtual hand modeling is one fundamental part of the model-based hand tracking system. The virtual hand model-based tracking system can be valuable in the field of the immersive human machine interaction simulation such as virtual surgery and virtual maintenance training. In this paper, a method of automatic stereo vision-based virtual hand modeling system is presented. We firstly built a template of the improved virtual hand model, then extracted the hand features from images captured by the binocular stereo vision system, and completed modeling the virtual hand in accordance with the estimated parameters. We propose a new method of combining local distance maximum and curvature vector to extract the hand features and introduce a method of recognizing the different finger tips. The advantages of feature extraction method loosen the high quality of contour extraction from images and enforce the robustness of feature extraction.