基于“原型”的机器人开放式室内场所感知实验研究  

Experimental Study on Open Interior-Places Perception of Robot Based on “Prototype”

作　　者：朱博[1] 戴先中[1] 李新德[1] 杨伟[1] 陈芳园[1] 

机构地区：[1]复杂工程系统测量与控制教育部重点实验室(东南大学自动化学院),江苏南京210096

出　　处：《机器人》2013年第4期491-499,512,共10页Robot

基　　金：国家自然科学基金资助项目(60804063;61175091);江苏省自然科学基金资助项目(BK2010403);图像信息处理与智能控制教育部重点实验室开放基金资助项目(200902);东南大学优秀青年教师教学;科研资助计划资助项目(3208001203);东南大学创新基金资助项目(3208000501)

摘　　要：在基于"原型"的场所感知算法基础上构建了一种场所感知实验平台.其中,硬件系统主要以双目立体相机作为环境感知传感器,以CPU-GPU(中央处理器–图形处理器)协同计算单元作为多种任务、复杂逻辑和密集数据的处理单元;而软件系统基于微软Robotics Developer Studio平台构建,双目视觉服务和场所感知服务为其中两大核心服务,前者利用RANSAC(随机采样一致性)和GPU版本ASIFT(仿射尺度不变特征变换)等算法在宽视角范围内实现物品识别和位姿估算,适用于拥挤室内环境,后者基于双目视觉结果和已有场所感知算法完成场所分类及区域感知并构建2D语义地图.在无人工标注的真实室内环境中开展实验,实验结果表明,基于本文实验平台的机器人能够在线鲁棒地感知活动空间中的开放式场所,实验也在一定程度上验证了已有场所感知算法的可实现性、有效性和实用性.A place-perception experimental platform is proposed on the basis of the prototype based place perception algorithm.In the hardware system,a binocular rig is used as environment perception sensor,and CPU（central processing unit） and GPU（graphic processing unit） coordinated computing unit is used to process multiple tasks,complex logic,intensive data,and so on.The software system is constructed based on Microsoft Robotics Developer Studio platform.Binocular vision service and place perception service are two core services in it.The former is fit for clustered indoor environment,and realizes object recognition and pose estimation simultaneously in wide viewing angle by using RANSAC（RANdom SAmple Consensus）,ASIFT（affine scale-invariant feature transform）-GPU and other algorithms.The latter realizes place categorization and region perception and constructs 2D semantic map based on the results of binocular vision and the existing place perception algorithm.The experiments are conducted in a real interior environment in which no artificial label is used.The experiment results show that the robot can robustly perceive open places in its action space online based on the experimental platform,and realizability,effectiveness and practicability of the existing place perception algorithm are verified to a certain extent.

关 键 词：场所分类 开放式室内场所 原型理论 场所感知 GPU ASIFT 语义地图 

分 类 号：TP242[自动化与计算机技术—检测技术与自动化装置]