An efficient high dynamic range 3D shape reconstruction method based on double phase-shifting profilometry

Abstract

Phase-shifting profilometry is the most widely used wrapped phase extraction method due to the advantages of high accuracy and resolution, robust to noise. However, due to the local specular reflection of the measured surfaces and the limited camera dynamic range, some pixels of the captured fringes become intensity saturated, thus introducing wrapped phase error. In this work, we simulate the Fourier spectrum of the saturated fringe and represent the saturated fringe approximately by a third-order Fourier series. Subsequently, we establish the saturation-induced wrapped phase error model. According to this model, it can be found that for the N-step phase-shifting method, the spatial frequency of the saturation-induced wrapped phase error is N times of that of the ideal wrapped phase. We phase shift the N-step phase-shifting fringe sequence π/N to generate the additional phase-shifting fringe sequence, and thus generate the opposite saturation-induced wrapped phase error. Finally, the saturation-induced wrapped phase error can be greatly reduced by fusing the wrapped phases of the two sets of phase-shifting fringe sequences. Compared with the traditional method and the adaptive projection fringe method, the phase error of the double N-step phase-shifting method is reduced by 67.63% and 65.2%, respectively.