Feature-based motion stereo matching

Abstract

Recovering scenes depth is important to robot vision. Stereo vision provides a convenient and direct way to obtain 3-D depth. The key problem in stereo vision is stereo image matching. In this paper a feature-based motion stereo matching algorithm is proposed. The stereo images are obtained by translating camera. The feature used is line sagments and three stereo images are used for line segments matching. The experiments with real images are made which show the effectiveness of our algorithm.