3D pointcloud processing using artificial intelligence and high performance computing techniques

Abstract

This thesis studies 3D point cloud processing as a challenge for computer science. High-performance computing techniques and artificial intelligence models are explored as the main strategies for efficiently processing 3D point clouds. Some tasks approached in this thesis are terrain segmentation, unsupervised clustering, analysis of crossing zones in urban contexts, classification of curb points and their interpolation, boundary segmentation, classical machine learning methods for point-wise classification, deep-learning models for point-wise classification and roughness estimation, and road bump detection through surface fitting method