Research On High-efficiency Stereo Matching Method Based On Fringe Projection Profilometry

The fringe projection profilometry(FPP)with accurate phase information to assist in stereo matching can achieve high-precision,high-density three dimensional(3D)reconstruction,which is widely used in industrial design,intelligent manufacturing,cultural heritage protection,medical detection and other fields.However,with the development of 3D applications,higher measurement requirements have been imposed on improving the measuring efficiency.At present,the main step that affects the measuring efficiency in the FPP is the phase stereo matching.This thesis mainly discussed stereo matching technique,and proposed the efficient stereo matching method to improve FPP performance.The main contents of the thesis include:(1)Aiming at the time-consuming problem caused by a large number of searches due to the large search range of corresponding points in the current stereo matching method,we propose a new way for reducing the search range of the corresponding points.The proposed method combines a camera and a projector in the binocular system to form a phase mapping system.Using an efficient phase-3D mapping calculation,the 3D point of the point to be matched is quickly estimated and projected onto another camera to obtain an estimated corresponding point.Theoretically,the corresponding point should coincide with the exact corresponding point,but due to the measurement error,there is a slight deviation between the two points.The deviation size is the corresponding point search range of the proposed method.Through experimental verification,the proposed method can reduce the search range of about 400 pixels within the parallax scope to 3 pixels,which greatly reduces the number of corresponding point search times,and improves the overall speed of stereo matching by about 4 times,thereby accelerating the 3D reconstruction rate.(2)To further improve 3D measurement speed by reducing FPP image sequence acquisition time,we propose an efficient method of stereo matching with the phaseunwrapping operation omitted.In this work,the bi-cameras system is utilized.We use the projector in the system as a reverse camera and use the geometric constraints between the two cameras and the projector to eliminate the matching ambiguity caused by the wrapped phase to achieve stereo matching.Finally we use the final matching results to reconstruct the 3D model.The proposed method simplifies the original algorithm flow by combining with the phase-3D mapping principle.By experimental verification,the method can reconstruct a 3D model with a degree of integrity higher than 97% with three phase-shift patterns,and has a greater improvement in efficiency than the original algorithm.