融合MeanShift和改进SURF算法的目标定位策略  

Targeting Strategy Based on MeanShift and Improved SURF Algorithm

作　　者：张毅[1] 张瀚 韩晓园 ZHANG Yi;ZHANG Han;HAN Xiao-yuan(Chongqing University of Posts and Telecommunications,Chongqing Information Accessibility and Service Robot Engineering Technology Research Center,Chongqing 400065,China)

机构地区：[1]重庆邮电大学重庆市信息无障碍与服务机器人工程技术研究中心,四川重庆400065

出　　处：《控制工程》2020年第4期629-634,共6页Control Engineering of China

基　　金：国家自然科学基金青年基金项目(61703067);重庆市教育委员会科学技术研究项目(KJ1704072)。

摘　　要：在工业环境中,针对工业机器人进行抓取物品操作时定位耗时长,精度低的问题,提出一种融合MeanShift算法和改进SURF算法的分层式定位策略,快速精确地定位物体。首先根据双目视觉系统采集的目标物体图片信息,采用MeanShift算法进行初步处理,切割出目标图片信息;然后利用改进后SURF算法对目标区域进行特征点对的匹配以及筛选;最后将匹配好的特征点对根据三角形测量原理,实现物体三维坐标的精确定位。实验验证了在工业机器人抓取物品时,本文所提方法对物体定位速度与精度上有所提升。In the industrial environment,a hierarchical object localization strategy combining Mean Shift algorithm and improved SURF algorithm is proposed to locate the object quickly and accurately when the object is manipulated by the industrial robot.Firstly,according to the target object image information collected by the binocular vision system,the MeanShift algorithm is used for cutting out the target image information for initial processing.Then,the feature points are matched by the improved SURF algorithm with good computational superiority and repeatability,uniqueness and robustness.Finally,the matched feature points are combined with triangular measurement principle to accurately locate the three-dimensional coordinates of the object.The experiment verifies that the method improves the speed and accuracy of object positioning when industrial robots grasp objects.

关 键 词：MEANSHIFT算法 改进SURF算法 特征匹配 三角形测量原理 双目视觉系统 

分 类 号：TP18[自动化与计算机技术—控制理论与控制工程]