HPC Solutions for ALS Point Cloud Processing in Pathfinding and Powerline Detection and Characterization

Abstract

This thesis addresses the processing of LiDAR point clouds using high-performance computing techniques. By employing efficient data structures and the shared-memory parallelization paradigm, two methods have been implemented for point cloud analysis. First, a path planning algorithm is used to find the route between any two points within an airborne LiDAR point cloud, considering terrain features such as trafficability, slope, roughness, presence of vegetation, and roads. It is guaranteed that the found route is optimal in terms of cost. Second, the problem of detecting and characterizing powerlines in general-purpose airborne LiDAR point clouds has been tackled. The method can detect multiple powerlines in a given scene with a precision of 97.2%, and it can model the conductors with a mean error of 0.14 meters.