在回环检测过程中，实时计算3D点云匹配，对算法的性能有极大的考验，算法

In the loopback detection process, real-time calculation of 3D point cloud matching has a great test on the performance of the algorithm, and the real-time performance of the algorithm cannot be guaranteed. Generally speaking, the recall rate of loopback detection can be sacrificed slightly, and the accuracy rate must be guaranteed. Therefore, in the loop detection of the fusion of lidar and RGB-D camera point cloud, the strategy adopted is to first use the 2D laser submap to find the appropriate loop, and after the loop is detected, the 3D point cloud descriptor is used for matching and verification The authenticity of the loopback, thereby improving accuracy.

In the process of loopback detection, real-time calculation of 3D point cloud matching is a great test of the performance of the algorithm, and the real-time performance of the algorithm can not be guaranteed. In general, the recall rate of the loop test can be slightly sacrificial, and the correct rate must be guaranteed. Therefore, in the return ring detection of fusion lidar and RGB-D camera point clouds, the strategy adopted is to use 2D laser submap to find the appropriate loop back, after the loop is detected, and then through the description of the 3D point cloud to match, verify the authenticity of the loop, and improve accuracy.

In the loop detection process, real-time calculation of 3D point cloud matching has a great test on the performance of the algorithm, and the real-time performance of the algorithm can not be guaranteed. Generally speaking, the recall rate of loopback detection can be slightly sacrificed, while the accuracy must be guaranteed. Therefore, in the loop detection of fusion lidar and rgb-d camera point cloud, the strategy is to use 2D laser submap to find the appropriate loop, after the loop is detected, then through the 3D point cloud descriptor to verify the loop authenticity, so as to improve the accuracy.