摘要
在已知镜面和透视相机参数的情况下,提出一种不需要任何其他标定物的折反射全向相机镜面位姿的自标定算法,只需要折反射相机采集任意一幅图像即可估计出反射镜面与透视相机之间的旋转和平移。利用镜面外边缘所成的像,通过平面圆位姿估算方法获得两个候选位姿;再利用透视相机镜头边缘的成像,同时进行镜头边缘参数估计和镜面位姿选择。该标定方法操作简单,精度高,适用于非单视点相机的标定。仿真和真图实验结果证明了该方法的有效性。
Given the parameters of the mirror shape and the camera,a self-calibration method for catadioptric camera system is proposed to estimate the mirror pose with respect to the camera by using only one image containing no special designed calibration pattern.The ellipse corresponding to the mirror boundary in the image is used to compute possible mirror pose by planar circular posture estimation.With two pose candidates thereafter,the correct pose and the parameter of lens boundary simultaneously is found by matching the predicted and the observed image of lens boundary.The method features easy-operating,high accuracy and is suitable for non-single viewpoint system.Experimental results both on synthesized and real images demonstrate the validity of our method.
出处
《光学学报》
EI
CAS
CSCD
北大核心
2012年第4期161-168,共8页
Acta Optica Sinica
基金
国家自然科学基金(61071219)
浙江省科技计划(2009C33118)资助课题
关键词
机器视觉
相机标定
镜面位姿
折反射相机
machine vision
camera calibration
mirror pose
catadioptric camera