Research on SLAM toolbox mapping based on LiDAR-camera fusion

To address the limitations of inspection robots that use SLAM Toolbox with lidar alone, such as incomplete obstacle detection or a restricted detection range, this paper proposes a SLAM Toolbox algorithm that fuses LIDAR and camera data. First, the initial point clouds from both the camera and lidar are pre-processed. The camera point cloud then undergoes voxel filtering before being converted into a 2D pseudo-laser point cloud, which is then fused with the lidar point cloud. To enhance odometry accuracy, an Extended Kalman Filter (EKF) is used to fuse IMU and wheel odometry data. Finally, this proposed method is validated on a dedicated robot platform. The experimental results show that the SLAM Toolbox, after fusing camera and lidar information, is able to build a grid map that better reflects the real-world factory environment. This approach improves the obstacle detection rate by 18.78 percentage points. Additionally, the EKF fusion of IMU and wheel odometry results in a 0.93 percentage point improvement in odometry accuracy compared to using wheel odometry alone.