基于多相机系统的高精度标定  

High-Precision Calibration Based on Multi-Camera System

作　　者：肖一帆 胡伟[1] Xiao Yifan;Hu Wei(School of Electrical Engineering and Automation,Henan Polytechnic University,Henan 454000,Jiaozuo,China)

机构地区：[1]河南理工大学电气工程与自动化学院,河南焦作454000

出　　处：《激光与光电子学进展》2023年第20期126-132,共7页Laser & Optoelectronics Progress

基　　金：国家自然科学基金(U1804147)。

摘　　要：随着工业机器视觉的深入发展,大视野高精度视觉系统的需求越来越多。针对大视野导致的精度过低问题,提出一种基于多个低像素相机联合标定的方法。在多相机中选择一个相机作为主相机,求取其他相机的像素坐标系与主相机像素坐标系的映射矩阵,使得主相机的视野无限扩展。同时,为了更精确地得到标定板图像中的圆心像素坐标位置,采用两步标定法提升标定精度。提取标定板圆心像素坐标进行第1次粗标定,获取相机内参以及标定板位姿,从而获取图像平面与世界坐标系的平面Z=0之间的映射关系。对其进行透视偏差矫正,提取矫正后标定板的圆心,再利用逆映射变换把相应的圆心转换回原始位置,用转换后圆心像素坐标位置对主相机进行第2次精标定。最后通过LevenbergMarquardt算法进行非线性优化获取全局最优解。实验结果表明,所提标定方法的重投影误差在0.005 pixel~0.01 pixel之间。Large-view,high-precision vision systems are becoming more and more in demand as industrial machine vision advances.Aiming at the issue of low accuracy caused by a large field of view,a joint calibration approach based on multiple low-pixel cameras is proposed.In a multi-camera setup,choose one camera to serve as the main camera.Then,determine the mapping matrix between the pixel coordinate systems of the other cameras and the main camera's pixel coordinate system such that the main camera's field of vision can be infinitely expanded.At the same time,to obtain the coordinates of the center pixel in the calibration plate image more precisely,the two-step calibration approach is employed to enhance the calibration accuracy.Extract the pixel coordinates of the center of the calibration plate for the first rough calibration,obtain the camera internal parameters and the calibration plate position and pose,and thus obtain the mapping correlation between the image plane and the plane Z=0 of the world coordinate system.After adjusting the perspective deviation,extract the center of the calibrated plate,use an inverse mapping transformation to return the associated center to its original location,and then calibrate the primary camera a second time using the converted center pixel coordinate position.Finally,the Levenberg-Marquardt algorithm is used for nonlinear optimization to achieve the global optimal solution.The experimental findings demonstrate that the re-projection error of the suggested calibration method is between 0.005 pixel-0.01 pixel.

关 键 词：机器视觉 低像素相机联合 透视偏差矫正 非线性优化 

分 类 号：TP391.41[自动化与计算机技术—计算机应用技术]