Banca de DEFESA: ALEXANDRE CURVELO DE ANDRADE

Uma banca de DEFESA de MESTRADO foi cadastrada pelo programa.
DISCENTE : ALEXANDRE CURVELO DE ANDRADE
DATA : 26/11/2019
HORA: 14:00
LOCAL: PROGRAMA DE PÓS-GRADUAÇAO EM ENGENHARIA CIVIL
TÍTULO:

 COMPARATIVE ANALYSIS OF NON-PARAMETRIC REGION CLASSIFICATION METHODS OF ROADS WITH HIGH RESOLUTION IMAGES AND LASER SCANNING

PALAVRAS-CHAVES:

Engineering. Cartographic engineering. Digital Photogrammetry and Remote Sensing. Digital Image Processing. Road Mesh Extraction. SVM. Artificial neural networks.

PÁGINAS: 120
GRANDE ÁREA: Engenharias
ÁREA: Engenharia Civil
RESUMO:

The present research is exploratory, regarding its objectives; of the Applied type as to its nature; of the Quantitative type, regarding the approach of the problem; and of the Bibliographic and Experimental type, regarding the technical procedures. The search method employed is the deductive one. The objective of the project is to comparatively analyze two non-parametric classification methods (the “Support Vector Machine” - SVM and the “Artificial Neural Network” - ANN) of regions of high spatial resolution images associated with Laser Scanning data. airborne. Similarly, the study aims to verify what kind of influence the attribute layers have on the performance of the respective classifiers (SVM AND ANN). The cartographic object that delimits the study is the databases related to the road network, as these are very important for human society due to their relevant role in people's daily lives. In addition, in recent decades, many methodologies have been proposed regarding the challenge posed by semi-automated road network extraction using remote sensing techniques. This difficulty can be proven in the review of some works related to the theme and published in the last decade. Mena (2003) and Ziems et al (2017) have shown that the problem of semiautomated extraction of road regions (streets or highways) is widely analyzed by various areas of science, and proposals based on genetic algorithms or expert systems are increasingly recurrent. In this context, nonparametric methods emerge as a strong trend. Therefore, the analyzes proposed in this project will be done in a controlled testing and validation environment, where both classifiers will receive the same set of training samples, the same sets of attributes obtained by aerophotogrammetry and with very high spatial resolution and radiometry ( gray level images of the visible spectrum, near infrared spectrum and Airborne Laser Scanning (ALS), as well as validated from the same true field image and through literature-consolidated procedures such as the Kappa Coefficient. A simple and non-repetitive combination of available information plans will allow us to analyze the influence of each layer on the attribute space and on the performance of each of the classifiers.

MEMBROS DA BANCA:
Externo à Instituição - EDSON APARECIDO MITISHITA
Interno - 2390672 - FERNANDA PUGA SANTOS CARVALHO
Presidente - 2466669 - MAURO JOSE ALIXANDRINI JUNIOR