大场景中基于IHDT的激光回环检测算法

IHDT-based laser loop closure detection algorithm in largescenes

针对大场景下机器人定位和地图构建中的累积漂移偏差问题,提出一种基于强度、高度与密度三进制模式(IHDT)描述子的激光回环检测算法。将点云空间划分为三进制单元块提取全局描述子IHDT,其中包含点云的强度、高度和密度信息。为提高回环检测的效率,避免直接匹配大场景中的关键帧,采用两阶段搜索算法;为确定与当前帧IHDT描述子最为相似的候选帧,通过检索历史帧IHDT的二进制矩阵,进而确定最终的目标回环索引。在KITTI 05大场景中,所提IHDT-LIO-SAM算法的轨迹均方根误差和标准差分别为1.81、2.23 m,相较于基于欧式距离的LIO-SAM回环检测算法分别减少34%和50%,显著降低了误差。在拥有多个回环的大场景中,所提方法展现出了更好的适应性。

Aiming at the problem of cumulative drift bias in robot localisation and map construction in large scenes,this study proposes a laser loopback detection algorithm based on the Intensity,Height and Density Trinary pattern(IHDT)descriptor.The algorithm divides the point cloud space into trinary cell blocks to extract the global description of IHDT,which contains the intensity,height and density information of the point cloud.In order to improve the efficiency of loopback detection and avoid direct matching of key frames in the large scene,a two-stage search algorithm was used.In order to determine the candidate frame that was most similar to the IHDT descriptor of the current frame,the binary matrix of the IHDT of the historical frames was retrieved,which in turn determines the final target loopback index.The root mean square error and standard deviation of the trajectory of the IHDT-LIO-SAM algorithm in the KITTI 05 large-scale scenario were 1.81、2.23 m,respectively,which were significantly reduced by the proposed method in this study compared to the Euclidean distance-based LIO-SAM loopback detection algorithm,with a reduction of 34%and 50%,respectively.The experimental data showed that the method in this study exhibits better adaptability in large scenarios with multiple loopbacks.