Robust and accuracy calibration method for a binocular camera using a coding planar target.

Abstract

High-accuracy binocular camera calibration is a vital basis of precise binocular vision 3D measurement. In this work, a high-precision and robust binocular camera calibration method based on a coding target is proposed. First, a coding target with the simple patterns is designed. Every corner on the coding target has a unique code number, which can make the identification of homonymous corners easier and more valuable, even if the target is partially occluded. The decoding of the coding target is rapid, robust, and accurate at a complex background. Subsequently, the zenith and azimuth angles are introduced in the proposed calibration method to study the effect of the orientation of the coding target on the stability of calibration results and improve the robustness of the calibration results. Finally, to fully utilize the 3D information of the calibration corners on the coding target, we combine the reprojection and 3D geometric constraints to propose a multi-constraint optimization method for refining the parameters of binocular camera and improving the accuracy of binocular camera calibration. The comparison experiments have been done to verify the performance of the proposed calibration method. The standard deviations of the intrinsic and extrinsic parameters are greatly decreased, compared with Zhang's method. The mean reprojection and 3D geometric errors calculated by the proposed method have a large reduction. And the application experiment furtherly validates the effectiveness of the proposed method.