基于改进椭圆拟合算法的钢管轮廓测量方法  

Profile measurement method of steel pipe based on improved ellipse fitting algorithm

作　　者：熊华德 张勇军[1] 石杰[2] 杨朝霖[1] XIONG Huade;ZHANG Yongjun;SHI Jie;YANG Chaolin(National Engineering Research Center for Advanced Rolling and Intelligent Manufacturing,University of Science and Technology Beijing,Beijing 100083,China;National Engineering Technology Research Center of Flat Rolling Equipment,University of Science and Technology Beijing,Beijing 100083,China)

机构地区：[1]北京科技大学高效轧制与智能制造国家工程研究中心,北京100083 [2]北京科技大学国家板带生产先进装备工程技术研究中心,北京100083

出　　处：《冶金自动化》2023年第4期109-116,共8页Metallurgical Industry Automation

摘　　要：基于机器视觉在线检测技术,采用椭圆拟合原理对钢管轮廓外形进行测量是一种有效的方法。针对传统椭圆拟合中离群点多所导致的拟合精度低和实时性差等问题,提出了一种改进算法,在截断最小二乘法的椭圆拟合算法基础上,设计了一种椭圆拟合的改进算法。该算法首先根据输入样本点的数量进行排序分段和移除,然后利用超过半数的点集进行椭圆参数的计算和迭代,直至找到拟合误差最小时对应的椭圆参数。通过在典型钢管实测的外轮廓数据上开展对比测试,结果表明,本文所提出的改进算法在样本点数量较多和存在离群点的情况下,可以获得更为理想的拟合精度与拟合速度,具有参数数量少、对自身参数不敏感等特点。Based on the machine vision online detection technology,it is an effective method to measure the profile of the steel pipe using the ellipse fitting principle.An improved algorithm was proposed to solve the problems of low fitting accuracy and poor real-time performance caused by many outliers in the traditional ellipse fitting.Based on the ellipse fitting algorithm of the truncated least square method,an improved algorithm for ellipse fitting was designed.Firstly,the algorithm performs sorting segmentation and removal according to the number of input sample points.Then,more than half of the points are used to calculate and iterate the ellipse parameters until the corresponding ellipse parameters with the smallest fitting error are found.Through comparative testing on the measured outer profile data of typical steel pipes,the results show that the improved algorithm proposed in this paper can obtain more ideal fitting accuracy and fitting speed when there are a lot of sample points and outliers,and it has the characteristics of fewer parameters and insensitivity to its own parameters.

关 键 词：钢管 轮廓测量 机器视觉 椭圆拟合 分段移除 

分 类 号：TG806[金属学及工艺—公差测量技术] TP391.41[自动化与计算机技术—计算机应用技术]