Feature-Based Mapping Method for Outdoor Complexity

Outdoor complex environments are characterized by intricate surface contours of objects and indistinct feature points within the scene. The uncertain number of feature points often leads to significant degradation in mapping quality and odometry drift. Based on LIO-SAM, we propose a SLAM method for localization in outdoor complex environments that relies on feature point extraction. Firstly, IMU pre-integration is used for LiDAR de-distortion. The processed point cloud undergoes adaptive curvature calculation to determine the curvature of each point. Plane feature points and corner points are then extracted based on their curvature values. Additionally, the local descriptor FPFH is introduced in loop closure detection, effectively enhancing the efficiency and robustness of point cloud registration. Ultimately, a high-precision point cloud map of outdoor complex environments is generated. Results on the KITTI dataset demonstrate that our method outperforms the LIO-SAM solution, reducing the absolute error trajectory by 29%, which meets the accuracy requirements of odometry.