Stratified Self-calibration and Metric Reconstruction of a Trinocular Structured Light Vision System

Abstract

A trinocular structured light vision system consists of a projector and two cameras. The using of two cameras instead of only one camera is to handle erroneous matching caused by reflection and blurring. A trinocular active vision system is useful in 3D reconstruction for its robustness and accuracy. One of the most difficult tasks in using such a system is the self-calibration of the system. In this paper a self-calibration and metric reconstruction algorithm for a trinocular structured light vision system is proposed. It is shown that the vanishing lines of the stripe planes in the camera view can be identified from the homographys induced by the planes and the geometry of the projective pattern. With the knowledge of the plane at infinity, the intrinsic and extrinsic parameters of both the cameras and the projector can be identified. Experimental results with real images are presented, demonstrating the accuracy and robustness of the proposed algorithm.