基于点云强度特征建图算法的轨道交通场景地图构建

Application of Mapping Algorithm Based on Point Cloud Intensity Characteristics to Map Construction in Rail Transit Scenarios

载体定位是实现自动驾驶的关键技术之一,其中基于地图的定位技术具有精度高和鲁棒性强的优点。同步定位与建图(simultaneous localization and mapping,SLAM)技术是针对未知环境地图构建的典型方法,在地铁隧道场景中,传统SLAM算法易因几何结构的严重退化而导致建图失败。针对此问题,文章提出一种基于点云强度特征的建图算法。其首先进行基于点云强度的特征点云提取,并根据广义迭代最近点算法完成高强度特征点云残差构建,以增加运动约束;其次,基于位姿图优化融合激光雷达数据与惯性测量单元数据,完成位姿优化与地图构建;最后,使用地铁隧道离线数据进行算法验证,结果显示,采用该方法成功构建了地铁全线的点云地图,地图无明显漂移,并采用隧道壁上固定安装距离的标识物体进行地图精度评估,所建地图平均偏差小于0.2m,验证了所提算法的有效性与鲁棒性。

Carrier localization is one of key technologies in the field of autonomous driving,among which map-based localization techniques have the advantages of high accuracy and robustness.Simultaneous localization and mapping(SLAM)technology serves as a typical method for map construction in unknown environments.However,in metro tunnel scenarios,the traditional SLAM algorithm often results in severe degradation in geometric structures,leading to unsuccessful map construction.To solve this issue,this paper proposes a mapping algorithm based on the intensity characteristics of point clouds.Firstly,feature point clouds were extracted based on point cloud intensity.Moreover,a generalized iterative closest point matching algorithm was introduced to construct residuals for high-intensity feature point clouds,thereby adding motion constraints.Secondly,pose graph optimization was fused with LiDAR data and IMU data to enable pose optimization and map construction.Finally,the offline data collected from real metro tunnels were used to verify the effect of the proposed algorithm,resulting in the successful construction of point cloud maps that cover entire metro lines,without significant drift.Map accuracy was evaluated using identifying objects at fixed installation intervals on the tunnel walls,demonstrating the algorithm's effectiveness and robustness,with an average deviation in maps of less than 0.2 m.