Spectral characterization and semantic segmentation of complex 3D models in natural environments

Résumé

This doctoral thesis presents main contributions in the characterization of 3D scenarios from multispectral images and the semantic segmentation of point clouds. As a result of this research, the GEU software (Geospatial and Environmental tools of University of Jaén) has been developed, which integrates the set of methods developed. This doctoral thesis is presented as a compendium of publications. The first part of this dissertation thesis focuses on the fusion of spectral and morphological variables of tree structures which have been reconstructed from the real world. An olive plantation and a forest area are used as study areas. The second part focuses on the analysis of several characteristics intrinsic to 3D models in order to propose advances in the semantic segmentation of different types of natural materials and the recognition of geometric patterns in vineyards.