This paper addresses outdoor terrain mapping using overhead images obtained from an unmanned aerial vehicle. Dense depth estimation from aerial images during flight is challenging. While feature-based localization and mapping techniques can deliver real-time odometry and sparse points reconstruction, a dense environment model is generally recovered offline with significant computation and storage. This paper develops a joint 2D-3D learning approach to reconstruct local meshes at each camera keyframe, which can be assembled into a global environment model. Each local mesh is initialized from sparse depth measurements. We associate image features with the mesh vertices through camera projection and apply graph convolution to refine the mesh vertices based on joint 2-D reprojected depth and 3-D mesh supervision. Quantitative and qualitative evaluations using real aerial images show the potential of our method to support environmental monitoring and surveillance applications.
updated: Tue Apr 13 2021 20:45:33 GMT+0000 (UTC)
published: Wed Jan 06 2021 02:09:03 GMT+0000 (UTC)