摘要
人眼难以明确自动驾驶车辆自身位置与驾驶环境,缺失场景同步数据,定位轨迹与实际轨迹拟合度较低,由此,提出基于三维激光点云的动态道路场景自动驾驶车辆定位方法。利用三维激光点云技术扫描动态道路场景,生成多图元,将三维激光点云数据投射至通视性较高的单视图中,利用任意二维函数提取单视图特征,根据地图元素的数据编译函数和动态道路场景耦合结果,反复联通真实场景,大面积复杂视图改良后实现定位。测试结果表明,车辆定位轨迹与实际轨迹拟合度较高,重构视差值最高为0.35,定位准确度高。
It is difficult for the human eye to determine the self position and driving environment of the autonomous driving vehicle,the scene synchronization data is missing,and the fitting degree between the positioning track and the actual track is low.Therefore,an automatic driving vehicle positioning method based on 3D laser point cloud in dy-namic road scene is proposed.The 3D laser point cloud technology is used to scan the dynamic road scene,generate multiple entities,project the 3D laser point cloud data into a single view with high visibility,extract the features of a single view using any two-dimensional function,compile the function and dynamic road scene coupling results accord-ing to the data of map elements,repeatedly connect the real scene,and realize positioning after improving the large ar-ea complex view.The test results show that the vehicle positioning track has a high fitting degree with the actual track,the maximum reconstructed parallax value is 0.35,and the positioning accuracy is high.
作者
王诚林
夏秀华
郭士茹
王蕾
季文超
高俊鹏
WANG Chenglin;XIA Xiuhua;GUO Shiru;WANG Lei;JI Wenchao;GAO Junpeng(Changchun College Of Electronic Technology,Changchun 130000,China;Mudanjiang Normal University,Mudanjiang Heilongjiang 157012,China)
出处
《激光杂志》
CAS
北大核心
2023年第11期188-193,共6页
Laser Journal
基金
吉林省高教科研课题(No.JGJX2021D501)。
关键词
三维激光点云技术
自动驾驶车辆
单视图
改进双边滤波
立体视觉
three-dimensional laser point cloud technology
autonomous driving vehicles
single view
improve bilateral filtering
stereovision
