基于双一维靶标的摄像机标定方法  

A new camera calibration method based on two 1-D targets

作　　者：杨珍[1] 孙军华[1] 吴子彦[1] 张广军[1] 

机构地区：[1]北京航空航天大学仪器科学与光电工程学院,精密光机电一体化技术教育部重点实验室,北京100191

出　　处：《光电子．激光》2010年第3期411-414,共4页Journal of Optoelectronics·Laser

基　　金：国家自然科学基金资助项目(60804060,50727502);高等学校博士学科点专项科研基金资助项目(200800061003)

摘　　要：针对大视场摄像机的标定,提出一种基于双一维靶标的摄像机标定方法。两个一维靶标任意放置在摄像机视场内,摄像机从多角度拍摄靶标图像;利用一维靶标特征点共线的特性进行图像畸变校正;计算两个一维靶标所在空间直线在图像中的消影点,得到在空间中的夹角关于摄像机内参数矩阵的表达式,利用其所成角度恒定这一约束求解摄像机内部参数;采用非线性优化方法,求解摄像机内、外参数在最大似然准则下的最优解。仿真试验结果表明,当噪声水平小于0.5 pixel时,摄像机内部参数的相对误差均小于0.3%。实物试验结果表明,视场范围为1 200 mm×800 mm时,该方法的标定精度优于0.1 mm。A new camera calibration method based on two 1-D targets is proposed to solve the calibration problem of camera with large field of view,which only requires the camera to observe two arbitrarily placed 1-D targets from several different viewpoints.First,the target images are undistorted using the co-linearity of the feature points on 1-D target.Then the vanishing point of the spatial line defined by each 1-D target is computed,and the angle between two 1-D targets is expressed using the camera intrinsic parameter matrix.The intrinsic parameters are solved using the invariability of angle between two 1-D targets in different views.Global optimization is implemented to obtain the maximum likelihood estimation of camera intrinsic and extrinsic parameters.Synthetic experiment suggests that the average errors of intrinsic parameters are within 0.3% when the noise level is under 0.5 pixel.Real calibration experiment shows that the projection error on the test plane is less than 0.1 mm in the 1 200 mm×800 mm field of view.

关 键 词：摄像机标定 大视场 一维靶标 消影点 

分 类 号：TP391[自动化与计算机技术—计算机应用技术]