Global Calibration of Virtual Multi-camera Vision System Based on 3D Target

Yang Liu, Haishu Tan, Fuqiang Zhou, Wanning Zhang

Abstract: Virtual multi-camera vision system (MVS) combining a single camera and catadioptric mirrors has been paid increasing attention for 3D measurement applications due to its wide field of view (FOV), low cost, and synchronization. Setups of the vision system require the global transformation relationships between the involved virtual cameras to be known. This work proposed a novel global calibration approach of virtual multi-camera vision system based on an unconstrained 3D target. Measurement model of the local virtual binocular structure and the global MVS as well as the corresponding parameter estimation methods are presented. Then the global parameters are refined by a non-linear optimization technique based on the reprojection matrix of adjacent local virtual cameras. Results of simulations validate the measurement model and the optimization method, showing that the proposed approach has the advantages of high accuracy, noise immunity and great flexibility.