摘要
针对线结构光视觉测量系统中参数标定易受噪声和亮度突变等因素的影响,而传统算法难以拟合包含异常值点云的问题,在现有标定方法的基础上,提出了一种基于随机抽样一致性(RANSAC)的改进算法拟合光平面参数。随机选取三个点云数据拟合光平面,选择阈值并统计在此平面上的内点数量,多次重复求得包含最多内点的平面,并以这些内点以特征值法进行平面拟合得到所求平面方程。实验结果表明,与最小二乘法相比,该方法可以很好地适应标定过程中出现的误差和异常值的情况,稳健地估计平面参数值,从而进一步提高线结构光参数标定精度。
Considering the calibration of parameters in line structured light vision measurement system is easily affected by the noise and brightness change and other factors,and it is difficult to fit outliers with classical algorithm. Based on the existing calibration methods,an improved algorithm of RANSAC is proposed to fit the light plane. Three points are randomly chosen to fit the light plane,then the threshold is selected and the number of inliers on this plane are counted,Finally the plane with the most inliers is obtained repeatedly,and the plane equations are obtained by fitting the inliers with the eigenvalue method. Experimental results show that compared with the least square method,this method can well adapt various situation of error and exception in the calibration process,and the parameters of the plane are estimated steadily. Thus,the calibration accuracy of line structured light parameters is further improved.
出处
《科学技术与工程》
北大核心
2018年第3期68-73,共6页
Science Technology and Engineering
基金
河北省自然科学基金(E2016202297)资助
