面向高动态环境的激光惯性SLAM算法  

LiDAR inertial SLAM for high dynamic environment

作　　者：周海霞 ZHOU Haixia(Shanxi Institute of Surveying and Mapping Geographic Information,Taiyuan 030001,China)

机构地区：[1]山西省测绘地理信息院,山西太原030001

出　　处：《经纬天地》2023年第2期97-100,共4页Survey World

摘　　要：定位与地图创建技术是智能驾驶和移动机器人的关键技术和重要前提,然而大多数的激光SLAM都基于静态世界的假设,导致现有算法难以在高度动态的场景中鲁棒地工作。为此,本文提出了一种基于聚类分割的激光惯性SLAM方法,通过对点云进行地面分割、聚类和动态分数计算,实现对动态目标的实时识别和剔除。本文在动态数据集Urbanloco上进行了广泛的实验。实验结果表明,该方法可以在动态场景中检测和去除运动目标,提高SLAM系统的定位精度和鲁棒性。Simultaneous localization and mapping is the key technology and important premise of intelligent driving and mobile robot.However,most LiDAR SLAM are based on static world assumptions that make it difficult for existing algorithms to work robustly in highly dynamic scenarios.Therefore,this paper proposes a LiDAR inertial SLAM method to achieve real-time recognition and rejection of dynamic targets by ground segmentation,clustering and dynamic score calculation on point clouds.In this paper,a wide range of experiments are designed and conducted on Urbanloco.The experimental results show that the method can detect and remove moving targets in the dynamic scenes and improve the localization accuracy and robustness of SLAM systems.

关 键 词：定位与地图创建 动态场景 激光惯性SLAM 

分 类 号：P237[天文地球—摄影测量与遥感]