Stereo match based on linear feature

Abstract

Techniques of image matching for stereo vision, generated with the aid of a structured light projector, are described, and are used to match primitives. Based on the description function of the primitive, the match rule is designed and four match constraints are used for eliminating the false match. A global to local control strategy is designed to accelerate the match process. This technique is used to obtain 3D measurements of human body, and to establish the surface representation for a computer-aided garment-design system. The results obtained indicate that this technique could also be used to acquire the 3D information for industrial parts and natural objects with smooth surfaces.<>