基于椭圆检测的小型工件分拣系统设计  

Design of Small Workpiece Sorting System Based on Ellipse Detection

作　　者：张庆新[1] 谢文傲 陈新禹[1] 庞济宝 牟金星 宫伟迪 Zhang Qingxin;Xie Wenao;Chen Xinyu;Pang Jibao;Mu Jinxing;Gong Weidi(School of Artificial Intelligence,Shenyang Aerospace University,Shenyang 110136,China;不详)

机构地区：[1]沈阳航空航天大学人工智能学院,沈阳市110136 [2]沈阳航空航天大学自动化学院

出　　处：《工具技术》2023年第1期156-160,共5页Tool Engineering

基　　金：辽宁省自然科学基金(JYT19002);辽宁省教育厅基础研究项目(JYT2020018)。

摘　　要：针对分拣系统中散乱堆叠的圆形工件检测和位姿估计问题,设计了基于椭圆检测的双目立体视觉小型工件分拣系统。使用LSD算法提取圆弧支撑线段,经筛选和连接组合后,采用改进的椭圆最小二乘拟合方法直接计算出初始椭圆;对初始椭圆进行层次聚类并筛选,完成椭圆识别。提取特征点,再根据Rosin近似距离筛选角点对左右目椭圆进行匹配并进行三维重建,根据随机抽样一致算法进行平面拟合,并提取圆形工件位姿。根据此结果应用Robot Studio控制机械臂移动和分拣工件。经测试和实验证明,该分拣系统能够迅速和准确地识别工件,完成自动分拣,在工件散乱堆叠的情况下,准确度较传统模板匹配方法大大提高。Aiming at the detection and pose estimation of scattered and stacked circular workpieces in the sorting system,a binocular stereo vision small workpiece sorting system based on ellipse detection is designed.The arc-support line segment is extracted by LSD,after the screening,connection and combination of line segments,the initial ellipse is calculated by an improved direct least square fitting method.The initial ellipse set is hierarchically clustered and screened to complete ellipse recognition.Then extracting feature points,screening corner points according to the approximate distance of Rosin,matching the left and right camera ellipses,performing three-dimensional reconstruction,and the plane fitting are carried out according to the random sampling consistency to extract the pose of circular workpiece.Finally,the manipulator controlled by Robot Studio is used to move and grab the workpiece for verification.Tests and experiments show that the sorting system can quickly and accurately identify the workpiece and automatically sorting.Compared with the traditional template matching method,it greatly improves the accuracy when the workpieces are scattered and stacked.

关 键 词：立体视觉 椭圆检测 分拣系统 

分 类 号：TG806[金属学及工艺—公差测量技术] TP242.2[自动化与计算机技术—检测技术与自动化装置]