Research on Lidar Vision Data Fusion Algorithm Based on Improved ORB-SLAM2

To address the issue of feature point loss in pure visual SLAM (Simultaneous Localization and Mapping) in complex environments, which leads to localization failure and poor mapping performance, this paper proposes an improved ORBSLAM2 algorithm by integrating IMU (inertial measurement unit) and wheel encoder data through EKF (Extended Kalman Filter) to enhance visual odometry. Additionally, Bayesian estimation is employed to fuse the improved algorithm with LiDAR SLAM for mapping. Simulation results demonstrate complete trajectory tracking throughout the experiment. In real-world testing, the fused algorithm generates more comprehensive maps compared to individual SLAM systems. Experimental results verify the feasibility of both the improved visual SLAM algorithm and the fusion algorithm for localization and mapping in complex scenarios.