Digital image correlation (DIC) based stereo matching method for binocular structured light system (BSLS)

With the advantages of non-contact, quick and high accuracy, binocular stereo vision technology is popular in the fields of industrial inspection and measurement. To improve the result of stereo matching, phase consistency constrain based on the fringe projection profilometry (FPP) is performed. The phase unwrapping is generally employed to avoid the phase ambiguity, which is unrobust or time consuming. Aiming at this problem, a digital image correlation (DIC) assisted phase consistency method is proposed to achieve stereo matching with high accuracy, only three fringe patterns and one digital speckle pattern are needed. Two-step strategy is performed to get the homonymy points. The epipolar constraint and DIC algorithm can get the matching with pixel level, and then the wrapping consistency constraint is used to get a sub-pixel matching. To improve the matching accuracy, the Hilbert transform is employed to compensate the phase nonlinear error. As to the regions with low modulation, the disparity refinement algorithm based on neighboring disparity constrain is performed. The experiment results show that the reconstruction accuracy of proposed method is comparative with the multi-step phase shift plus multi-frequency heterodyne method.