基于双目立体视觉测量数据的空间圆拟合方法对比

Comparison of spatial circle fitting methods based on measuring data collected from binocular stereo vision system

为了找出测量工程应用中最优的空间圆拟合方法,进行了基于双目立体视觉测量数据的空间圆拟合方法对比.根据相交成空间圆的平面和球面的求解次序及关联性,将现有5种拟合方法分为平面与球面共同求解、平面与球面独立求解、平面与球面关联性求解3类;利用自行研制的双目立体视觉系统,采集3组随多齿分度台做圆周运动的靶标的空间坐标作为拟合数据;以心面距、平面度、圆度为拟合精度评价指标,通过空间圆拟合试验,对比分析各拟合方法的性能及影响因素.试验结果表明:球面需要平面约束才能正确求解;先利用最小二乘平差法求解平面法向量和常系数,再利用坐标系变换法或中垂面法求解球心和半径是最优拟合方法;增加测量点数能够提高拟合精度.

To find the optimal spatial circle fitting method in the measuring engineering application, the comparison of spatial circle fitting methods was completed based on the data collected from a binocular stereo vision system. According to the solving order and relevance of plane and spherical surface intersecting into a spatial circle, the existing five fitting methods were divided into three categories of solving plane and sphere together, solving plane and sphere independently and solving plane and sphere correlatively. The spatial circle fitting experiment was carried out to compare and analyze the performance and influence factors of fitting methods. Using the selfdeveloped binocular stereo vision system, three sets of spatial coordinates of a target making circular motion with multitooth dividing table were collected as fitting data, and the centerplane distance, the flatness and the roundness were used as fitting accuracy evaluation indexes. The experimental results show that the spherical surface needs the plane constraint to be solved correctly. The optimal fitting method is that the normal vector and constant coefficient of the plane are firstly solved by the leastsquare adjustment method, and then the center and radius of the sphere are solved by the coordinate system transformation or the perpendicular bisection plane method. The fitting accuracy can be improved by increasing the number of measurement points.