基于双目立体视觉的偏转线圈测量方法研究

立体匹配过程中,由于视差搜索范围不易确定,使匹配极易出现错误.文中提出了一种灰度相关多峰值视差绝对值极小化立体匹配新方法.通过对最大灰度相关匹配法的分析,指出最大灰度相关系数所对应的点并不完全对应于正确匹配,由此提出了灰度相关匹配多峰值法,以获得初始匹配集.结合视差梯度约束,对所得初始匹配集进行部分优化,采用视差绝对值极小化确定唯一匹配点.通过对一形状复杂且不规则物体(偏转线圈)及已知三维坐标的标准件的三维重建,证实了文中所提方法的有效性,可满足三维不规则物体(偏转线圈)三维空间坐标的非接触精密测量要求.

A Study of Measurement on Deflection Yoke Based on Binocular Stereo Vision

The matching result is very easy to be wrong due to the uncertainty of searching scope on disparity during stereo matching.A novel stereo matching algorithm is developed,whose criterion is the minimum of absolute disparity combined with gray correlation multi peak value method.It is pointed out that the corresponding point with maximum of gray normalized correlation coefficient is not the correct matching point completely by analyzing maximum of gray normalized correlation coefficient matching approach,so gray correlation multi peak value matching method is mentioned to obtain initial matching aggregation in the paper.Disparity gradient restriction is used to optimize the initial matching aggregation partially.The method,whose criterion is the minimum of absolute disparity,is used to ascertain the only matching point.It is proved that the algorithm proposed in the paper is feasible by 3D reconstruction on an object (deflection yoke,DY)with complex and irregular surface,and a standard piece with given 3D coordinates,which can meet the demands for measuring in non contact the 3D coordinates of object,such as DY,with irregular surface precisely.