基于单幅图片的相机完全标定  

Full Calibration of Camera Based on Single Image

作　　者：李竹良[1] 赵宇明[1] 

机构地区：[1]上海交通大学自动化系系统控制与信息处理教育部重点实验室,上海200240

出　　处：《计算机工程》2013年第11期5-8,共4页Computer Engineering

基　　金：国家自然科学基金资助项目"多视角下的多类型目标识别与行为分析"(61175009)

摘　　要：现有相机标定方法的标定过程比较繁琐,不利于标定相机的广泛使用。为此,从摄像机镜头畸变矫正着手,结合标定板信息及消失点约束,提出一种基于单张图片的相机标定方法。利用非线性迭代得到相机镜头的畸变系数,通过线性求解得出相机的内参,直接计算得到相机的外参,从而实现仅需拍摄单张标定板图片的相机完全标定。实验结果表明,该方法在标定板与视平面夹角小于45的情况下均能成功标定,并且重投影误差小于0.3像素。The popular calibration method is relatively a complex stuff, which obstructs the privilege of calibrated cameras. Based on the lens undistortion method, auxiliary with calibration board information and constrains from vanish points, this paper proposes a camera full calibration method based ＇on single image. The lens distortion parameters are acquired with Levenberg-Marquardt（LM） iteration. Subsequently, camera＇s intrinsic parameters are linearly solved. Camera＇s extrinsic parameters are gained directly. The proposed method is easy to use for only one image contains a calibration board is needed to reach the full calibration. Experiments are employed to certify that it is robust enough to achieve calibration when the angle between calibration board and view plane is less than 45 degree, meanwhile keeping a low re-projection error which is less than 0.3 pixels.

关 键 词：相机标定 畸变矫正 快速标定 标定板 消失点 针孔相机 

分 类 号：TP317.4[自动化与计算机技术—计算机软件与理论]