A High-Precision Calibration and Optimization Method for Stereo Vision System

Abstract

Stereo-vision plays an important part in planetary exploration. A high-precision calibration method is introduced for such system, which uses wide-angle lens. It makes use of planar homography constraint to estimate intrinsic and extrinsic parameters. On the basis of this initial guess, an optimization scheme is used to minimize a new cost function, 3D reprojection error sum. All Calibration parameters of two cameras are globally optimized simultaneously with Genetic Algorithm (GA). Both simulation and real image experiment results show this scheme have much higher precision than traditional methods.