Occlusion detection by surface-gradient-based method for true-orthophoto generation

Author(s)

Henrique Cândido de Oliveira

• Document type: Doctoral Thesis
• Press: Presidente Prudente. 2016-06-22.
• Institution: Universidade Estadual Paulista (Unesp). Faculdade de Ciências e Tecnologia. Presidente Prudente
• Defense date: 2016-04-29
• Advisor: Aluir Porfírio Dal Poz; Mauricio Galo

Abstract

The usage of true-orthophoto in projects performed over urban areas is essential for certain applications. To generate this type of product it is necessary a procedure for occlusion detection, as well as, a proper radiometric compensation for the occluded areas, using adjacent images from the aerial image block. The main objective of this thesis is to present a new method for occlusion detection named SGBM (Surface-Gradient-Based Method), using a surface gradient identification over a polyhedral surface. The main difference among this method and the others, presented as state of the art, is the usage of a point cloud arranged as a triangular data structure (TIN: Triangulated Irregular Network). This surface representation is formed by an irregularly-spaced 3D point cloud, which allows the use of different point density along the surface. This characteristic supports obtaining good results by using a less dense point cloud - situation that requires a reduced computational effort. Several experiments were carried out, seeking for the validation of the proposed method, by using quantitative analyzes (completeness index, mosaic alignment and time of processing) and qualitative analyzes (visual coherence of the obtained results). Three datasets were used, being all acquired by a low-altitude flight configuration (helicopter and unmanned aerial vehicles), which make possible the identification of large occlusion areas. The point cloud was obtained by airborne LASER (Light Amplification by Stimulated Emission of Radiation) scanning system and by dense image matching. These dataset were used due to the incidence of occlusions and the growth of unmanned platform usage. The results indicate that the proposed method has some contribution to the state of the art in occlusion detection, aiming the generation of true-orthophoto. The use of an irregularly-spaced point cloud avoids some limitations found on the other methodologies, along with, a better representation of objects in the surface of interest. (AU)

• FAPESP's process: 13/21647-2 - THE USAGE OF LASER DATA FOR OCCLUSION DETERMINATION USING HEIGHT GRADIENT FROM A POLYHEDRIC SURFACE, FOR TRUE-ORTHOPHOTO GENERATION.
• Grantee: Henrique Cândido de Oliveira
• Support Opportunities: Scholarships in Brazil - Doctorate