An effective framework for 3D shape measurement of specular surface based on the dichromatic reflection model

Abstract Fringe projection profilometry based on phase-shift method has been commonly used for three-dimensional measurement. However, measuring the specular surface is always a challenging task since the specular area cannot present original intensity information. This paper presents a novel framework for measuring objects with specular surfaces. Firstly, a set of phase-shifted fringe patterns with maximum intensity value is projected onto the specular surface. Then, the dichromatic reflection model is developed to theoretically decompose an image into specular and diffuse reflection components, which identifies specular pixels more robustly and effectively than judging by saturation only. Secondly, since the location and intensity of specular areas are acquired, the best projection intensity corresponding to the specular areas can be determined by fitting a power function that transforms captured specular reflection components to projected intensities, aiming to minimize specular areas and decrease specular reflection components. It should be stated here that the assumption based on this paper is that the fewer specular reflection components, the fewer measurement errors. Subsequently, the adaptive fringe patterns, which are constructed using the best projection intensity, are reprojected for phase recovery. Thirdly, it is reasonable that there still exist regions where the specular reflection components are too high. For these information-loss areas, the pixel inpainting technique is adopted when extracting the disparity map. Finally, the ultimate result combines adaptive fringe reconstruction with pixel inpainting. The experiment results verify that the proposed measurement framework can effectively and robustly perform three-dimensional reconstruction of various objects with specular surfaces.