Geometrically Constrained Phase Unwrapping Method Based On Variable-frequency Fringe Pattern And Neighborhood Feature Matching

Structured light measurement technology has the advantages of non-contact, full field, high robustness, high speed, and high precision, and has been widely used in industrial inspection, machine vision, cultural relics digitization, medicine, and other fields. Multi-view structured light method introduces an auxiliary camera on the basis of single-view structured light. According to the principle of geometric constraints phase unwrapping, the wrapped phase can be unwrapped without projecting additional fringe patterns. Multi-view structured light has become the focus of structured light measurement technology research in recent years because of its advantages of small number of projected fringe patterns, fast phase unwrapping, and high efficiency of point cloud reconstruction. However, the current multi-view structured light geometric constraints phase unwrapping algorithm still has the problems of high matching difficulty and low phase unwrapping accuracy under the high-frequency fringes. Addressing these problems, this paper proposes a geometric constraints phase unwrapping method based on variable-frequency fringe patterns and neighborhood features matching. The method calculates neighborhood features of the pixel points in left camera wrapped phase image and the candidate points in right camera wrapped phase image. Then, the best matching point among the candidate points is obtained by using the difference of neighborhood features of the pixel points with the same phase in the different frequency fringe. Therefore, the pixel points matching between the left camera view and right camera view can be implemented, and the geometric constraints phase unwrapping can be realized. The proposed method does not need to project additional fringe patterns. It can thus reduce the time cost of projecting patterns, capturing images, and unwrapping phases. In addition, the proposed method can increase the difference of neighborhood features of the candidate points with the same phase in the different frequency fringe, and thus improve the reliability of pixel points matching and the accuracy of geometric constraint unwrapping. Experimental results show that this method can effectively reconstruct the 3D point cloud of objects with spatial discontinuities and complex geometric shapes by only projecting four variable-frequency sinusoidal fringe patterns.