基于棋盘标定板的优化相机参数标定方法  

Optimized Camera Parameter Calibration MethodBased on Checkerboard

作　　者：林绿开 钮倩倩 李毅 LIN Lyu-kai;NIU Qian-qian;LI Yi(School of Computer and Artificial Intelligence,Wenzhou University,Wenzhou 325000,China;School of Computer Science and Technology,Zhejiang University,Hangzhou 310058,China)

机构地区：[1]温州大学计算机与人工智能学院,浙江温州325000 [2]浙江大学计算机科学与技术学院,浙江杭州310058

出　　处：《计算机技术与发展》2023年第12期101-105,共5页Computer Technology and Development

基　　金：浙江省软科学研究计划重点项目(2022C25033);温州市科技计划项目(R20200025)。

摘　　要：相机标定是计算机视觉与重建定位的重要基础,是构建二维图像与三维空间联系的重要桥梁。该文根据传统线性相机成像的规律,在理想的针孔相机成像模型的基础上,同时针对张正友标定法中相机参数求解过程较为复杂与繁琐的问题,采用一种简易的标定方法来求解单目相机的相机参数。该方法在平面张正友标定方法的基础上,通过预先估计部分相机参数的初值进而求解理想针孔相机成像模型后,再进一步优化预估参数的方式,根据棋盘标定板的特征点的空间坐标与图像坐标的对应方程,利用奇异值分解和Levenberg-Marquardt优化的方法进行参数求解,从而避免求解繁杂的约束矩阵后再分别对相机参数进行求解的过程,达到直接求解相机参数的目的,优化求解过程,最终实现单目相机的相机标定。实验结果表明,使用该方法标定的单目相机在物距一米以下时的标定误差能够满足绝大部分视觉与定位的应用需求,且该方法具有较好的稳定性和可行性。Camera calibration is an essential basis for computer vision and reconstruction positioning,which is a critical bridge between two-dimensional images and three-dimensional space.According to the regulation of traditional linear camera imaging and the ideal pinhole camera imaging model,a simple calibration method is proposed to solve the camera parameters of the monocular camera,which can address the complex and tedious calculation problem based on the plane Zhengyou Zhang Calibration Method.The method is based on the plane Zhengyou Zhang Calibration Method,which optimizes the predicted parameters by pre-estimating the initial values of some camera parameters and solving the ideal pinhole camera imaging model.According to the equations of the spatial and image coordinates of the checkerboard calibration board's feature points,the singular value decomposition and Levenberg-Marquardt optimization methods are used for parameter solution,which avoids the process of calculating the complicated constraint matrix before the camera parameters separately,thus achieving the purpose of solving the camera parameters directly,while optimizing the solving process and finally realizing camera calibration of monocular cameras.The experimental results show that the calibration error of the monocular camera calibrated with the proposed method can satisfy most vision and positioning applications at object distances below one meter,with high stability and feasibility.

关 键 词：单目相机 标定算法 奇异值分解 Levenberg-Marquardt优化 张正友标定法 

分 类 号：TP391.41[自动化与计算机技术—计算机应用技术]