Optimization of Calibration Method for Binocular Fisheye Camera and Application of Stereo Matching

A camera calibration method combining radially symmetric model with tangential distortion is described, proposing that the tangential distortion of the fisheye camera model should be calculated after the radial distortion, instead of the simple linear summation. The method adds the de-tangential distortion process in the local fine optimization of the external parameters, and uses the Newtonian iterative method to estimate the camera coordinates after de-tangential distortion. On this basis, a binocular fisheye camera system is established and stereo calibrated. The stereo calibration results are used to de-distort and stereo match the images, and compare the traditional stereo matching with the stereo matching algorithm based on RAFT-Stereo network. Finally, the camera calibration and stereo matching are compared and experimented to analyze the calibration reprojection error and matching error, and the experimental results verify the accuracy and feasibility of the method.