Incremental Triangle Mesh Generation with Mesh Refinement

This letter presents an incremental algorithm to generate triangle meshes from Light Detection and Ranging (LiDAR) point clouds with mesh refinement. The algorithm produces triangle mesh directly from LiDAR output without storing a dense point cloud to create a high-quality triangle mesh. In our algorithm, as the number of captured points increases during the LiDAR operation and robot movement, the new scan points from the LiDAR output refine or extend the existing triangle mesh. Such an approach is suitable for computationally-constrained systems like aerial vehicles, mobile robots, and smartphones, as it requires relatively limited resources. Our algorithm can reconstruct the topology of city- sized scenes maintaining a maximum triangle mesh error below 2 cm much faster than state-of-the-art triangle mesh generation algorithms that we demonstrate on publicly available data sets.