A Method for Mapping and Localization of Quadrotors for Inspection under Bridges Using Camera and 3D-LiDAR

Bridge inspection is necessary to maintain the safety of bridges built in many places. Recently, taking pictures or capturing 3D data of all parts of the bridge for inspection are performed using quadrotors instead of humans. Since manual control of the UAV needs time and human resources, researchers are developing localization methods under the bridge using a camera or a 3D LiDAR sensor. However, there are difficulties in localization under large bridges due to lack of features. In this paper, we present a novel method for mapping and localization under large bridges by combining visual inertial state estimate and 3D LiDAR data. Due to lack of features, we combine multiple state estimation like visual inertial odometry, GPS, and 3D LiDAR data into one supernode. Then NDT (Normal distributions transform) and Plane ICP (Iterative closest point) are performed to match each node and to optimize the path using constraints. Generated 3D point cloud map contains LiDAR reflection intensity and appropriate camera images to inspect the bridge more efficiently. The 3D point cloud map results are compared with other 3D point cloud maps generated using visual inertial odometry only.