Multisensor registration using phase matching for large-scale fringe projection profilometry

Abstract

Multisensor systems are widely used in large-scale fringe projection profilometry, where the accuracy and efficiency of registration play an important role. However, existing methods using reference targets or auxiliary devices lengthen the coordinate transformation chain and are difficult to achieve high accuracy and efficiency. In this study, we propose a method that uses the absolute phase retrieved from sinusoidal fringe patterns for the registration of different sensors, which can shorten the coordinate transformation chain. First, the corresponding pixels of cameras of different sensors are determined through phase matching and used to calculate the corresponding points on the basis of triangulation, which determines corresponding points between the sensors directly. The coordinate transformation is then calculated. Furthermore, the registration accuracy can be improved by combining the epipolar constraint and bundle adjustment. Experimental results show that the proposed method can be used for accurate and efficient registration without auxiliary devices and reference targets.