非结构化环境下基于外观的闭环检测研究综述  

A Survey of Appearance-based Loop Closure Detection in Unstructured Environment

作　　者：李康宇 王西峰 徐斌 姬丽娟 耿牛牛 LI Kangyu;WANG Xifeng;XU Bin;JI Lijuan;GENG Niuniu(China Academy of Machinery Science and Technology Group,Beijing 100037,China;Machinery Technology Development Co.Ltd.,Beijing 100037,China)

机构地区：[1]中国机械科学研究总院集团有限公司,北京100037 [2]机科发展科技股份有限公司,北京100037

出　　处：《机器人》2023年第2期238-256,共19页Robot

摘　　要：首先讨论了现有的场景外观描述方法及其应对各类非结构化因素时的性能。其次,介绍了基于纯图像检索、引入拓扑和度量信息的场景记忆模型的特性,并作为案例讨论了视觉词典的性能优化及构造方法。接着,对闭环检测的关键环节——闭环匹配、后验和优化——进行了分析。再次,概述了常用的性能评估指标和基准数据集。最后,总结了闭环检测研究现状,展望了无监督学习、语义上下文信息和模型轻量化等新技术的应用潜力。Firstly,the existing scene appearance description methods and their performances when dealing with various unstructured factors are discussed.Secondly,the characteristics of scene memory models based on pure image retrieval,and topological information and metric information are introduced,and as a case,the performance optimization and construction methods of the visual dictionary are discussed.Thirdly,three key steps of loop closure detection(LCD)are analyzed,including loop closure matching,subsequent verification,and optimization.Fourthly,the popular evaluation metrics and the standard datasets for LCD methods are presented.Finally,the current situation of LCD research is summarized,and the potential of novel techniques including unsupervised learning,semantic context,and lightweight model is described.

关 键 词：视觉SLAM 闭环检测 深度学习 位姿优化 

分 类 号：TP242[自动化与计算机技术—检测技术与自动化装置]