基于平面基线靶标的视觉测量数据拼接方法

VISION MEASUREMENT DATA REGISTRATION METHOD BASED ON PLANAR BASELINE TARGET

针对视觉测量中三维数据的拼接问题,提出一种基于平面基线靶标的数据拼接方法。该方法利用平面基线靶标提供特征点,建立视觉传感器在相邻位置坐标系旋转矩阵和平移矢量各参数的优化目标函数,通过Levenberg—Marquardt优化算法求解出旋转矩阵和带有未知系数平移矢量。在此基础上,利用平面靶标上的基线长度确定平移矢量系数,从而求得数据拼接矩阵。利用平面靶标上100个点进行拼接试验,计算出x、y和z坐标拼接RMS误差分别为0．038 mm、0．022 mm和0．135 mm。结果表明该方法柔性好,适用范围广。

A data registration method based on planar baseline target is proposed for the 3D data registration problem in vision measurement. An optimization object function about all parameters of rotation matrix and transmission vector between two vision sensor coordinate frames at adjacent viewpoints is established using feature points provided by the planar target. The rotation matrix and the transmission vector with an unknown scale are estimated by Levenberg-Marquardt optimization algorithm. The unknown scale is found relying on baseline length on the planar target, and the data registration matrix is obtained. The registration experiment is conducted using 100 feature points on the planar target, and the data registration RMS error ofx, y andz is 0.038 mm, 0.022 mm and 0.135 mm respectively. The experimental results show the presented method is flexible and widely applicable.