语义地图中语义物体免疫遮挡定位算法  

An Immune Occlusion Localization Algorithm for Semantic Objects in Semantic Map

作　　者：左建朋 蒋林[1,2] 李云飞 邵慧超 刘奇 ZUO Jianpeng;JIANG Lin;LI Yunfei;SHAO Huichao;LIU Qi(Key Laboratory of Metallurgical Equipment and Control Technology of Ministry of Education,Wuhan University of Science and Technology,Wuhan 430081,China;Institute of Robotics and Intelligent System,Wuhan University of Science and Technology,Wuhan 430081,China;Leador Spatial Information Technology Corporation,Wuhan 430073,China)

机构地区：[1]武汉科技大学冶金装备及其控制教育部重点实验室,武汉430081 [2]武汉科技大学机器人与智能系统研究院,武汉430081 [3]立得空间信息技术股份有限公司,武汉430073

出　　处：《组合机床与自动化加工技术》2023年第7期22-27,共6页Modular Machine Tool & Automatic Manufacturing Technique

基　　金：国家重点研发计划项目(2019YFB1310000)。

摘　　要：针对语义物体被遮挡而导致语义定位失效或存在多区域多位置定位位姿的问题,提出了一种在语义地图中免疫语义物体被遮挡的定位算法。首先,利用已经建立的二维语义地图生成语义链,当有障碍物遮挡语义物体时,对当前帧与构建语义地图时保存的关键帧进行基于词袋模型相似度匹配和关键帧语义链检验,确定正确相似帧;其次,进行帧间语义向量链加权求差值,根据差值确定被遮挡的语义物体类别;最后,根据当前帧包含的语义物体种类查语义地图的语义链来确定机器人所处的区域,并依据该区域的高置信度语义信息完成语义定位。通过在搭建的移动机器人平台进行定位实验验证,能快速和精确的进行重定位,相对AMCL算法迭代收敛速度提高了60%,证明本文所提算法的真实有效性。In order to solve the problem that the semantic localization fails or there are multi-region and multi-location positioning poses because of the semantic object occluded when running semantic localization algorithm In this paper,an immune semantic object occlusion localization algorithm in semantic map is proposed.First of all,the established 2D semantic map is used to generate semantic chain.When there is the obstacle block key semantic information,the similar frame is determined based on the similarity matching of the Bag-of-words model and the semantic chain matching for the current frame,and the semantic information category that is occluded is determined according to the difference value by the weight of the semantic vector chain between frames.Afterwards search the map semantic chain to determine the robot's area and the semantic prepositioning is completed according to the high confidence semantic information of the region.Finally,the positioning experiment is carried out on the mobile robot platform,which can quickly and accurately carry out the relocation and the iterative convergence speed is improved by 60%compared with the original AMCL,he validity of the proposed algorithm is proved.

关 键 词：语义信息 语义链 相似帧 置信度 遮挡免疫 

分 类 号：TH165[机械工程—机械制造及自动化] TG659[金属学及工艺—金属切削加工及机床]