基于标记点的机器人抓取目标期望点动态匹配  

Dynamic Matching of Desired Points for Robots Grabbing Targets Based on Markers

作　　者：张天益 张磊[1] 李红兵 王旭光 宗泽 ZHANG Tianyi;ZHANG Lei;LI Hongbing;WANG Xuguang;ZONG Ze(School of Mechanical Engineering,Nantong University,Nantong Jiangsu 226019,China;Department of Instrument Science and Engineering,Shanghai Jiao Tong University,Shanghai 200240,China)

机构地区：[1]南通大学机械工程学院,江苏南通226019 [2]上海交通大学仪器科学与工程系,上海200240

出　　处：《传感技术学报》2022年第2期207-216,共10页Chinese Journal of Sensors and Actuators

基　　金：国家自然科学基金项目(51505280);上海市自然科学基金项目(18ZR1421300)。

摘　　要：针对特征点匹配过程中非标准形状目标物提取的特征点具有随机性,无法对目标物上某一点持续提取;且受环境影响,特征点匹配准确率不高等问题,提出一种基于标记点的匹配方法。首先在目标物期望跟踪点附近贴上若干个标记点,再通过YOLOv3对各标记点进行识别框选;然后在提取区域内进行FAST特征点检测并选取Harris响应值最大的N个特征点进行BRISK描述并匹配;将匹配结果作为标记点的相似度描述得到各帧标记点的对应关系;将各区域内的特征点进行最小二乘法拟合椭圆求解标记点中心;最后根据标记点预设的几何关系和匹配关系实现对期望点的动态匹配。实验表明该方法相较于传统SIFT、Harris-SIFT、光流法,能在目标物背光、自然光干扰等复杂环境下对期望点实现较好的匹配效果。In the process of feature point matching,the feature points extracted from the non-standard shape target are random,and it is impossible to continuously extract a point on the target;and the accuracy of the feature point matching is not high due to environmental influences,a marker-based method is proposed.First,paste a number of marker points near the desired tracking point of the target,and then use YOLOv3 to identify each marker point;then perform FAST feature point detection in the extraction area and select the N feature points with the largest Harris response value for BRISK Describe and match;use the matching result as the similarity of the marker points to describe the corresponding relationship of the marker points in each frame;perform the least squares method to fit the feature points in each area to the ellipse to find the center of the marker point;finally,according to the preset geometry of the marker point Relationships and matching relationships achieve dynamic matching of desired points.Experiments show that compared with traditional SIFT,Harris-SIFT,and optical flow methods,this method can achieve a better matching effect to desired points in complex environments such as target backlighting and natural light interference.

关 键 词：计算机图象处理 特征点匹配 标记点 椭圆拟合 YOLOv3 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]