通常来说深度相机的探测距离有限，直接使用其点云数据会导致使用场景受限，

Generally speaking, the detection distance of the depth camera is limited. Direct use of its point cloud data will lead to limited use scenarios and can only be used in small scenes. The error of the point cloud increases sharply with the distance. The point cloud data at the edge The credibility is extremely low, and if it is directly matched, the confidence of the robot's posture is also relatively low. The detection distance of the lidar is much farther than the RGB-D camera, and the accuracy is extremely high, usually in the millimeter level. Therefore, this paper proposes a point cloud fusion strategy to fuse 2D laser point cloud and RGB-D camera point cloud.

Generally speaking, the depth camera's detection distance is limited, the direct use of its point cloud data will lead to limited use of scenes, can only be used in small scenes, and the error of point cloud with the distance increases sharply, the point cloud data at the edge of the confidence is very low, direct matching, the robot's position confidence is also low. Lidar is detected much far the distance compared to rGB-D cameras and is extremely accurate, usually in millimeters. Therefore, this paper proposes a point cloud fusion strategy, 2D laser point cloud and RGB-D camera point cloud fusion.

Generally speaking, the detection distance of the depth camera is limited, and the direct use of its point cloud data will lead to the limited use of the scene and can only be used in small scenes. Moreover, the error of the point cloud increases rapidly with the distance, and the credibility of the point cloud data at the edge is very low. If it is directly matched, the robot's pose confidence is also relatively low. The detection range of lidar is much longer than that of rgb-d camera, and the accuracy is very high, usually in millimeter level. Therefore, this paper proposes a point cloud fusion strategy, which fuses 2D laser point cloud and rgb-d camera point cloud.