Multiple map representations for vehicle localization and scene reconstruction

Mengyin Fu, Hao Zhu, Yi Yang, Meiling Wang, Yu Li

Abstract: This paper proposes a RBPF SLAM algorithm for localization map building, and the SLAM result can be used for fusing multiple 3D photorealistic local maps of the environment. For the localization map, a heightmap representation based on the Velodyne pointcloud is presented, and a RBPF SLAM algorithm is established for precise mapping and localization. For the scene reconstruction map, the Velodyne LIDAR and Ladybug3 omnidirectional camera are fused to provide a photorealistic 3D representation of the scene.