在障碍物检测时对斜坡点云的检测处理算法  

Detection and processing algorithm of slope point cloud in obstacle detectio

作　　者：蒋林[1,2] 杨立 张文俊 张琼玉 吴艳霞 JIANG Lin;YANG Li;ZHANG Wen-jun;ZHANG Qiong-yu;WU Yan-xia(Key Laboratory of Metallurgical Equipment and Control Technology,Ministry of Education,Wuhan University of Science and Technology,Wuhan 430081,China;Institute of Robotics and Intelligent Systems,Wuhan University of Science and Technology,Wuhan 430081,China;Leador Spatial Information Technology Co.,Ltd.,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：《吉林大学学报（工学版）》2023年第11期3221-3228,共8页Journal of Jilin University:Engineering and Technology Edition

基　　金：国家重点研发计划项目(2019YFB1310000);湖北省重点研发计划项目(2020BAB098);国家自然科学基金项目(51874217)。

摘　　要：室外移动机器人在包含斜坡路面的环境下进行障碍物检测时,采用传统的RANSAC地面点云剔除算法会出现无法剔除斜坡点云的现象,导致后续将斜坡识别成障碍物。针对这一情况,本文提出了一种相邻线束算法来检测和剔除斜坡点云,并基于欧式聚类对障碍物进行检测。该方案首先对三维激光点云进行预处理,根据三维激光雷达相邻线束之间的几何关系分割出包含斜坡的地面点云,再经过体素滤波进行下采样后,对分割出的障碍物点云进行基于KDTree的欧式聚类,并通过PCA主成分分析计算出外接包围框的尺寸和方向来对障碍物进行检测。经过实验表明:采用本文提出的障碍物检测算法能有效分割出环境中的斜坡路面,在聚类过程中能够避免将斜坡路面识别为障碍物,为机器人自主行走避障策略提供基础。When the outdoor mobile robot detects obstacles in the environment including sloping roads,the traditional RANSAC ground point cloud removal algorithm will be unable to remove the slope point clouds,resulting in the subsequent identification of the slope as obstacles.In view of this situation,this paper proposes an Adjacent Laser Points Algorithm to detect and eliminate slope point clouds,and detect obstacles based on Euclidean clustering.In this scheme,the 3D laser point cloud is preprocessed,and the ground point cloud including the slope is segmented according to the geometric relationship between the adjacent lines of the 3D lidar,and then downsampled by voxel filtering,the segmented obstacle point cloud is clustered based on KDTree,and the size and direction of the outer bounding frame are calculated by PCA principal component analysis to detect the obstacle.The experimental results show that the obstacle detection algorithm proposed in this paper can effectively segment the slope road surface in the environment,and can avoid identifying the slope road surface as an obstacle in the clustering process,which provides the basis for the robot autonomous walking obstacle avoidance strategy.

关 键 词：机械电子工程 障碍物检测 相邻线束 欧式聚类 斜坡路面 PCA主成分分析 

分 类 号：TP391.9[自动化与计算机技术—计算机应用技术]