基于IRANSAC-IRLS直线拟合算法及应用  

Line Fitting Algorithm Based on IRANSAC-IRLS and Its Application

作　　者：罗金鐄 胡小平[1] 彭向前 黄泓 LUO Jinhuang;HU Xiaoping;PENG Xiangqian;HUANG Hong(School of Mechanical Engineering,Hunan University of Science and Technology,Xiangtan 411201,China)

机构地区：[1]湖南科技大学机电工程学院,湘潭411201

出　　处：《自动化与仪表》2023年第9期87-91,共5页Automation & Instrumentation

基　　金：国家自然科学基金项目(61572185);湖南省教育厅重点资助项目(19A170)。

摘　　要：针对工业视觉检测中直线边缘存在沾连、毛刺等噪声,导致拟合效率不高、精度较差的问题,提出一种基于梯度方向改进的随机采样一致性(improved random sample consensus,IRANSAC)的迭代加权最小二乘(iterative reweighted least-squares,IRLS)直线拟合算法,即IRANSAC-IRLS算法。首先,利用直线上边缘点的梯度方向相近,将梯度方向引入边缘点RANSAC拟合,来降低错误的随机抽取的次数;然后,对IRANSAC提取出来的局内点进行迭代加权最小二乘拟合,求得最终的直线参数。在噪声点比例为20%、40%、60%、80%的条件下,将IRANSAC-IRLS与基于随机采样一致性算法的最小二乘(RANSAC-LS)拟合算法的仿真实验结果进行对比,IRANSAC-IRLS比RANSAC-LS的拟合效率分别提高16.3%、41.9%、47.5%、53.2%,拟合精度分别提升14.3%、16.7%、44.0%、69.0%。Aiming at the problems of low fitting efficiency and poor accuracy caused by noise such as adhesion and burr on straight edge in industrial visual inspection.An iterative weighted least-squares(IRLS)algorithm based on the random sample consensus algorithm modified by gradient direction(IRANSAC)is proposed,which called IRANSAC-IRLS algorithm.First of all,the gradient direction of the edge points on the line is similar,and the gradient direction is introduced into the RANSAC fitting of the edge points to reduce the number of wrong random samples.Then,iterative weighted least square fitting is performed on the local points extracted by IRANSAC to obtain the final line parameters.The simulation results show that,the IRANSAC-IRLS was compared with the least square method(RANSAC-LS)based on the random sampling consistency algorithm under the conditions of 20%,40%,60%and 80%noise ratio.The fitting efficiency of IRANSAC-IRLS was improved by 16.3%,41.9%,47.5%,53.2%respectively,and the fitting accuracy was also improved 14.3%,16.7%,44.0%,69.0%respectively,compared with RANSAC-LS.

关 键 词：直线拟合 梯度方向 随机采样一致性 迭代加权最小二乘法 

分 类 号：TP391[自动化与计算机技术—计算机应用技术]