摘要
针对64线和128线激光雷达价格过于昂贵,而16线激光雷达点云过于稀疏无法直接进行语义信息提取的问题,本文提出了一种组合导航辅助的激光雷达-相机实时语义建图方法,并通过算法结构设计保障了语义建图的实时性。首先,在组合导航定位结果辅助下,完成了不同采集时刻的点云-图像配准;其次,从图像目标检测框中准确提取了语义物体类别的点云。基于移动机器人平台测试评估了语义建图性能,结果表明该方法能够有效的提高语义点云提取的准确率并在嵌入式处理器Xavier上实时构建语义地图,为移动机器人利用语义信息进行导航和实时执行任务奠定了基础。
Aiming at the problem that 64-beam and 128-beam scanning LiDAR are too expensive,while point cloud of 16-beam scanning LiDAR is too sparse to extract semantic information directly,a method for real time semantic mapping based on Lidar-Camera system aided by integrated navigation is proposed in this paper.The algorithm structure is designed to ensure the real time performance of semantic mapping.Firstly,with the help of the positioning results of integrated navigation,the point cloud and image acquired at different times are registered;Secondly,the point cloud of semantic objects are accurately extracted from the object bounding boxes of image.The performance of semantic mapping is tested and evaluated based on mobile robot platform.The results show that the method can effectively improve the accuracy of semantic point cloud extraction,and build a semantic map in real time on embedded processor Xavier,which lays a foundation for mobile robot to use semantic information for navigation and real-time task execution.
作者
张乐翔
刘天弋
张提升
牛小骥
ZHANG Lexiang;LIU Tianyi;ZHANG Tisheng;NIU Xiaoji(GNSS Research Center,Wuhan University,Wuhan Hubei 430079,CAina)
出处
《传感技术学报》
CAS
CSCD
北大核心
2021年第3期297-304,共8页
Chinese Journal of Sensors and Actuators
基金
国家重点研发计划项目(2020YFB0505803)
国家自然科学基金项目(41974024)。
关键词
语义建图
激光雷达
相机
点云-图像配准
语义点云提取
组合导航
semantic mapping
Li DAR
camera
point cloud and image registered
semantic point cloud extraction
integrated navigation
