A new microscopic telecentric stereo vision system - Calibration, rectification, and three-dimensional reconstruction

In stereo vision based three-dimensional (3D) measurements, calibration and stereo matching are the most challenging tasks for accurate 3D reconstruction. In this paper, a new microscopic telecentric stereo vision system is proposed to retrieve 3D data of micro-level objects by direct triangulation from two accurately calibrated telecentric cameras. The complex projector calibration procedure commonly seen in traditional structured-light based systems can be avoided. Besides, an improved and practical calibration framework of telecentric cameras is presented. Compared with existing calibration methods which retrieve the distortion parameters without reasonable initial guesses, our proposed approach derives reliable initial guesses for all the camera parameters to avoid the local minima problem based on the estimation algorithms before iterations. To realize precise sub-pixel stereo matching, we propose an effective searching algorithm based on the epipolar rectification of the absolute phase maps obtained from fringe projection profilometry. Experimental results indicate that a measurable volume of 10 mm (L)  ×  7 mm (W)  ×  7 mm (H) is achieved with the standard deviation of 1.485 μm.