室内服务机器人实时目标识别与定位系统设计  

Real-time object recognition and localization system for indoor service robot

作　　者：黄海卫[1] 孔令成[2,3] 谭治英[2,3] 

机构地区：[1]常州大学信息科学与工程学院,江苏常州213164 [2]中国科学院合肥物质科学研究院,安徽合肥230031 [3]常州先进制造研究所,江苏常州213164

出　　处：《计算机工程与设计》2016年第8期2228-2232,共5页Computer Engineering and Design

基　　金：青年科学基金项目(61401437);中科院合肥研究院院长基金项目(YZJJ201322);江苏省科技支撑计划基金项目(SBE201201023)

摘　　要：针对服务机器人在室内复杂环境下难以准确获得物体标识和空间信息的问题,提出一种结合三维点云分割和三维全局特征匹配的实时物体识别与定位系统。对获取的点云数据进行快速有效分割,计算场景内最大水平面,通过改进的聚类算法提取水平面上的物体聚类,采用基于聚类视点特征直方图(CVFH)特征匹配的方法识别目标物体,计算目标所在空间位置。实验结果表明,该识别定位系统满足服务机器人自主实时准确抓取的要求。Aiming at the problem that service robots in complex environment are difficult to accurately obtain object identification and spatial information,the real-time object recognition and localization system was proposed which integrating 3Dpoint clouds segmentation and global feature matching.Segmentation was carried out effectively and efficiently to find the maximum horizontal surface of the primary point cloud data.The improved clustering algorithm was used to extract the objects,and objects were identified using the CVFH feature matching based approach.Experimental results show the system can well meet the real-time and accuracy requirements.

关 键 词：服务机器人 物体识别 分割 三维全局特征 聚类 

分 类 号：TP391.41[自动化与计算机技术—计算机应用技术]