Stereo Matching for Monotone Texture of Mechanical Parts

This article presents an algorithm complementing classical stereo matching technique for mechanical parts. Mechanical parts have special surface characterizations of having monotone texture and specular reflections. Most measurements of mechanical parts by photogrammetry often resort to structured light method. Some preprocessing is applied to alleviate specular reflections and enhance the image quality. The algorithm combines edge detection and area matching. Edge detection determines the disparity range and area matching provides dense disparity map under this guideline. The algorithm solves two problems on the mechanical parts surface, not only improves matching accuracy but also reduces searching range in each scan line.