Multimodal 3d Facade Reconstruction Using 3d Lidar And Images

Reconstruction of 3D facades is an important problem in systems that reconstruct urban scenes. Facade reconstruction can be challenging due to the typically large featureless surfaces involved. In this work, we investigate the use of combining a commercially available LiDAR with a GoPro camera to serve as inputs for a system that generates accurate 3D facade reconstructions. A key challenge is that 3D point clouds from LiDARs tend to be sparse. We propose to overcome this by the use of semantic information extracted from RGB images, along with a state-of-the-art depth completion method. Our results demonstrate that the proposed approach is capable of producing highly accurate 3D reconstructions of building facades that rival the current state-of-the-art.