摘要
采用分离式差分标定方法实现对单摄像机的高精度标定。在摄像机的物面位置上放置多个小尺度的分离式标定靶,每个标定靶上有标准点阵列,对每个光点由高斯曲面拟合法得到其CCD像面上的位置,光点位置检测结果的稳定性(均方根)达到1μm(相当于CCD 0.0053 pixel)。利用同一标定靶上标准点之间的物面差分坐标,进行四次曲面拟合,建立摄像机的像面与物面之间的映射关系。该方法对各个标定靶之间的距离没有提出要求,采用多个小尺寸的差分标定靶代替整体式的大尺度标定靶,降低对大尺度标定靶的制作要求。实验表明,当各个标定靶之间的相对角度误差小于60″时,标定残差平均值可以达到4.8μm(相当于CCD 0.028 pixel)。
The difference calibration and separation patterns are used to calibrate a single-camera with high precision. Several separation calibration patterns with small size are put on the object plane of the camera. Each pattern has some spot array with high precision. Gaussian curved-surface fitting method is used to get the position of spot on the CCD image plane, and the stability （RMS） of measurement result of the spot position is 1μm （about 0. 0053 CCD pixel）. The object plane difference coordinates of the reference points on the same calibration pattern are used to calibrate the camera. The mapping relationship between the object plane and the image plane is established with the biquadratic curved-surface fitting method. There is not need for the distance between the patterns. The integer pattern with large dimension is replaced by the several small difference patterns. The difficulty to manufacture the large pattern is avoided. When the error of the relative angle between the patterns is less than 60＂. The experiment results show that the calibrated mean value of residual error is 4.8μm （about 0. 028 CCD Dixels）.
出处
《光学学报》
EI
CAS
CSCD
北大核心
2006年第5期697-701,共5页
Acta Optica Sinica
关键词
机器视觉
摄像机标定
差分式标定
分离式标靶
曲面拟合
machine vision
camera calibration
difference calibration
separation pattern
curved-surface fitting
