A Dynamic SLAM Algorithm Based on Lidar-Vision Fusion

SLAM (simultaneous localization and mapping) is based on the assumption of a static environment, and the external sensors equipped with SLAM have their own advantages and disadvantages. Therefore, poor lighting conditions, lack of geometric features and dynamic objects in the scene will interfere with the positioning accuracy of the SLAM algorithm based on a single sensor. To solve this problem, this paper proposes a dynamic SLAM algorithm based on Lidar-vision fusion. The algorithm provides depth information for image features through the fusion of monocular image sequences and Lidar scans. After rough positioning with fusion data, dynamic objects in the scenes are eliminated to further optimize the positioning accuracy. The comparative experimental results evaluated on 11 sequences in the KITTI odometry datasets demonstrate that the localization accuracy of the proposed algorithm is better than the vision-based ORB-SLAM2 and DynaSLAM algorithms.