Urban reconstruction from a video captured by a surveying vehicle constitutesa core module of automated mapping. When computational power represents alimited resource and, a detailed map is not the primary goal, thereconstruction can be performed incrementally, from a monocular video, carvinga 3D Delaunay triangulation of sparse points; this allows online incrementalmapping for tasks such as traversability analysis or obstacle avoidance. Toexploit the sharp edges of urban landscape, we propose to use a Delaunaytriangulation of Edge-Points, which are the 3D points corresponding to imageedges. These points constrain the edges of the 3D Delaunay triangulation toreal-world edges. Besides the use of the Edge-Points, a second contribution ofthis paper is the Inverse Cone Heuristic that preemptively avoids the creationof artifacts in the reconstructed manifold surface. We force the reconstructionof a manifold surface since it makes it possible to apply computer graphics orphotometric refinement algorithms to the output mesh. We evaluated our approachon four real sequences of the public available KITTI dataset by comparing theincremental reconstruction against Velodyne measurements.
展开▼
