Binocular stereo vision calibration based on constrained sparse beam adjustment algorithm

Abstract In order to realize high-precision three-dimensional measurement of railway train head, a high-precision camera calibration algorithm for binocular stereo vision system based on constrained sparse beam adjustment is proposed in this paper. By analyzing the digital model of binocular vision and the principle of three-dimensional measurement, a binocular stereo vision measurement system based on time-domain coding is built. The edge preservation filtering algorithm based on mean square deviation is used to denoise and filter the visual image. The image is enhanced by histogram equalization. A FAST algorithm based on accelerated segmentation detection is proposed to extract feature information. To solve the problem of camera calibration, a binocular vision calibration method based on coplanar intersecting circle is proposed. The coordinates of circular points are calculated by using the extracted elliptic contour of the intersecting circle, and then the internal parameters of the two cameras and the position relationship between them are obtained. The calibration algorithm needs no manual intervention and has high accuracy. Aiming at the distortion correction, the camera calibration is transformed into a global optimization problem, and the sparse beam adjustment algorithm based on constraints is used to optimize the parameters. Compared with Zhang's calibration method, the results show that the average absolute relative error of the parameters in the two cameras is less than 3.45e−6, the average re-projection error is about 0.1 pixels, and the data accuracy is high, which proves the feasibility and theoretical correctness of the experimental method of the system.