面向复杂场景的地图轻量化与匹配定位方法  

Map lightweight and matching localization method for complex scene

作　　者：魏东岩 巨柳荫 纪新春 马朝伟 袁洪 WEI Dongyan;JU Liuyin;JI Xinchun;MA Chaowei;YUAN Hong(Aerospace Information Research Institute Chinese Academy of Science,Beijing 100094,China;School of Electronic,Electrical and Communication Engineering,University of Chinese Academy of Science,Beijing 101408,China;School of Electronics and Information,Northwestern Polytechnical University,Xi’an 710072,China)

机构地区：[1]中国科学院空天信息创新研究院,北京100094 [2]中国科学院大学电子电气与通信工程学院,北京101408 [3]西北工业大学电子信息学院,西安710072

出　　处：《中国惯性技术学报》2025年第2期114-123,共10页Journal of Chinese Inertial Technology

基　　金：国家自然科学基金青年科学基金(42204048)。

摘　　要：针对移动机器人在计算、存储与通信资源受限场景下高精度定位需求,提出一种环境特征地图轻量化重建与匹配定位方法。设计自适应点特征直方图(APFH)描述子表征点特征显著度,充分提取关键特征点构建轻量化点云地图,有效兼顾复杂场景中不同类型区域的地图轻量化与特征丰富性。基于粗-精级联匹配定位方法,使用采样一致性配准算法(SAC-IA)搜索匹配APFH完成粗匹配,并提出对称化广义迭代最近点算法(SGICP)进行精匹配,通过对称化构造误差函数降低匹配陷入局部最优概率,提高轻量化点云地图的匹配定位精度。公开数据集与实际场景测试表明,与现有常用方法相比,相同压缩率下所提方法构建的轻量化点云地图信息熵平均提高49.63%,多类场景下平均匹配定位精度提高17.19%。A lightweight reconstruction and matching localization method for environmental feature maps is proposed to meet the high-precision positioning requirements in scenarios with limited computing,storage,and communication resources for robots.An adaptive point feature histogram(APFH)is designed to describe the saliency of sub feature points,and the key feature points are fully extracted to construct a lightweight point cloud map,effectively balancing map lightweight and feature richness for different types of regions in complex scenes.Based on the coarse fine cascade matching localization method,the sampling consistency registration(SAC-IA)algorithm is used to search for matching APFH to complete coarse matching,and the symmetric generalized iterative nearest point algorithm(SGICP)is proposed for fine matching.By constructing a symmetric error function,the probability of falling into local optima is reduced,and the matching localization accuracy of lightweight point cloud maps is improved.Public datasets and actual scenario testing show that compared with existing methods,the information entropy of the lightweight point cloud map constructed by the proposed method under the same compression ratio is increased by 49.63%on average,and the average matching positioning accuracy in multiple scenarios is increased by 17.19%.

关 键 词：点云匹配 地图轻量化 对称化广义迭代最近点算法 自适应点特征直方图 

分 类 号：TP242[自动化与计算机技术—检测技术与自动化装置]