最后再将点云数据插入到submap，将激光雷达的数据进行重投影，构建最

Finally, the point cloud data is inserted into the submap, the lidar data is reprojected, the least square problem is constructed, and the matching accuracy is further improved by means of optimization. After the matching is completed, the current laser point and RGB-D camera point are inserted into the submap. When the number of frames reaches a certain size, it is saved as a key frame for subsequent loopback detection and the establishment of a new key frame.

Finally, the dot cloud data is inserted into the submap, the lidar data is re-projected, the least squares problem is constructed, and the matching accuracy is further improved by optimization. After the matching is complete, the current laser point and RGB-D camera point are inserted into the submap, and when the number of frames reaches a certain size, save as a key frame for subsequent loopback detection, while the new key frame is established.

Finally, the point cloud data is inserted into the submap, and the lidar data are re projected to construct the least square problem, and the matching accuracy is further improved by means of optimization. After matching, the current laser points and rgb-d camera points are inserted into the submap. When the number of frames reaches a certain scale, they are saved as key frames for subsequent loopback detection and the establishment of new key frames.<br>