P-LVIO: A Plane-Based Lidar Visual Inertial Odometry in Urban Environments

To enhance localization accuracy of unmanned systems in urban environments, a novel LiDAR, visual and inertial fused odometry algorithm based on plane feature in urban environment is proposed in this paper, named as P-LVIO. The plane feature is utilized to enhance the correlation accuracy between LiDAR and visual, thereby improving the localization accuracy of unmanned systems. The general steps of the proposed algorithm are outlined as follows: firstly, the pose of the unmanned system is estimated and the environmental map is constructed with LiDAR-Inertial odometry. The constructed environmental map is then divided into grids, organized by a Hash table. Next, plane feature is extracted from the grids and projected onto images, and the selected pixels within the projected area are used for optical flow tracking. Finally, the pose of unmanned systems is estimated by minimizing the reprojection error. The proposed method is validated using real-world measurements to demonstrate its effectiveness in both indoor and outdoor urban environments.