基于强度信息特征滤波的激光SLAM算法  

Intensity-based Feature Filtering for LiDAR-based SLAM

作　　者：李维刚 邹少峰[2] 王永强 余楚翔 Li Weigang;Zou Shaofeng;Wang Yongqiang;Yu Chuxiang(Engineering Research Center for Metallurgical Automation and Measurement Technology of Ministry of Education,WuhanUniversity of Science and Technology,Wuhan 430081,China;School of Information Science and Engineering,Wuhan University of Science and Technology,Wuhan 430081,China)

机构地区：[1]武汉科技大学冶金自动化与检测技术教育部工程研究中心,湖北武汉430081 [2]武汉科技大学信息科学与工程学院,湖北武汉430081

出　　处：《系统仿真学报》2025年第2期392-403,共12页Journal of System Simulation

基　　金：国家重点研发计划(2019YFB1310000);湖北省揭榜制科技项目(2020BED003);湖北省重点研发计划(2020BAB098)。

摘　　要：为解决过多的特征点参与点云配准导致算法精度下降和建图效果不佳问题,提出一种基于强度信息特征滤波的激光SLAM算法。根据点云强度信息计算局部地图中特征点附近的强度分布情况,并为局部地图中的每个特征点赋予一个强度分布指标;通过设定强度阈值,滤除在连续帧中强度信息变化较大的无效特征点,筛选出适于将扫描顿与局部地图进行配准的有效特征点。提出一种强度加权代价函数,以获取机器人当前帧在全局地图中更准确的位姿。仿真结果表明:相较PFilter算法,每一扫描顿中的特征点数量平均减少了13.8%,局部地图中的特征点数量平均减少了18.8%,配准精度提高了5.7%。In order to solve the problem that an excessive influx of feature points into the point cloud registration phase can potentially lead to diminished algorithmic accuracy and suboptimal mapping outcomes,a novel laser SLAM algorithm predicated on the filtering of feature points through the utilization of intensity information is proposed.The intensity distribution near the feature points in the local map is calculated based on the point cloud intensity information,and each feature point within the local map is attributed an intensity distribution index.Through the application of an intensity threshold,feature points that exhibit substantial variations in intensity across successive frames are systematically removed.This process identifies and retains only valid feature points amenable for the registration of scanned frames with the local maps.A novel intensity-weighted cost function is proposed.This function aims to enhance the accuracy of the robot frame's pose estimation within the global map.The simulation results show that it achieves an average reduction of 13.8%in feature point count per scan frame,along with an average decrease of 18.8%in the number of feature points within the local map,enhances registration accuracy by5.7%.

关 键 词：同步定位与建图 激光雷达 强度信息 特征滤波 点云配准 

分 类 号：TP242[自动化与计算机技术—检测技术与自动化装置]