基于改进MDGHM-SURF算法的混联式输送机构末端位姿检测  

Position and Attitude Measurement for a Hybrid Mechanism Based on Modified MDGHM-SURF

作　　者：高国琴[1] 刘梦茹 GAO Guoqin;LIU Mengru(School of Electrical&Information Engineering,Jiangsu University,Zhenjiang 212013)

机构地区：[1]江苏大学电气信息工程学院,镇江212013

出　　处：《计算机与数字工程》2018年第7期1433-1439,共7页Computer & Digital Engineering

基　　金：国家自然科学基金项目(编号:51375210);江苏高校优势学科建设工程项目(编号:苏政办法[2014]37号);镇江市工业科技支撑计划项目(编号:GY2013062);镇江市京口区科技计划项目(编号:jk GY2013002);江苏省高校自然科学基金项目(编号:15KJB510006)资助

摘　　要：针对该课题组研制的混联式输送机构运行过程中机构末端连接杆图像特征点少,易受其他构件遮挡难以准确检测的问题,提出了基于Hough-K均值改进MDGHM-SURF算法的混联式输送机构末端位姿检测方法。首先采用双向滤波改进的Canny算子提取图像边缘点,再对边缘点运用Hough变换提取直线边缘点,然后结合机构特征采用K均值算法对提取的直线边缘点进行筛选以获取机构特征点,并采用MDGHM特征描述符描述筛选后的特征点,最后匹配左右两幅图像中特征点的特征向量,使得匹配结果在提高准确性的基础上,兼具快速性和稳定性。实验结果表明,对于该输送机构,基于双目视觉,采用结合Hough-K均值的MDGHM-SURF匹配算法较常规SURF匹配算法提取机构特征点的准确率提高6%,进而使位姿检测的准确度得以提高。For a hybrid mechanism,aiming at the moving rod easily blocked by other mechanism parts which make accurate detection difficult and the characteristic points less,a method for the hybrid mechanism position and attitude measurement based on modified MDGHM-SURF algorithm are presented.Canny algorithm improved by bilateral filter is adopted to extract edge points,and Hough transformation is uesd to extract straight edge points.Combined with the mechanism feature,the characteristic points is classify selected by K-means algorithm,described by MDGHM descriptor and matched by the characteristic description vectors.The experiment results show that for the conveying mechanism,based on the binocular vision,in comparison to SURF algorithm,the MDGHM-SURF algorithm combined with Hough-K-means algorithm improves the accuracy of extracting the interest points by 6%,which can significantly improve the accuracy of detection.

关 键 词：输送机构 双目视觉 特征提取 图像匹配 位姿检测 

分 类 号：TP242.62[自动化与计算机技术—检测技术与自动化装置]