Open Journal Systems

OBJECT ORIENTED ANALYSIS AND SEMANTIC NETWORK FOR HIGH RESOLUTION IMAGE CLASSIFICATION

ALZIR FELIPPE B. ANTUNES, CHRISTEL LINGNAU, JORGE ANTONIO SILVA CENTENO

Abstract



Este artigo apresenta a classificação digital de imagem de alta resolução baseada em
orientação a objeto. Os objetos da imagem (Ikonos, multiespectral) são oriundos de
segmentação multi-resolução que permite a obtenção de diferentes níveis de
segmentação. A segmentação em diferentes camadas pode apoiar uma estrutura
hierárquica de segmentos, onde uma rede semântica pode ser introduzida. A rede
semântica é baseada no processo cognitivo do usuário. O conhecimento pode ser
explicitado por meio de regras fuzzy considerando descritores de forma e textura dos
objetos, bem como as suas relações topológicas - entre objetos e subobjetos. São
testados diferentes métodos de classificação dentro da rede semântica tais como
seletivo e mudança de contexto. A área teste é uma zona agrícola, no Município de
Nova Esperança-PR, onde se propôs mapear a mata ciliar ao longo do Rio Porecatú.


Abstract


This work presents a high resolution image classification based on object oriented.
The objects are derived from multiresolution segmentation ( from multispectral
image, Ikonos). It allows a creation of different levels of segments supporting a
hierarchical structure, generating spatial relations between objects and sub-objects.
This hierarchy is the bedding for the semantic network. The knowledge is the
semantic basis. The classification is based on fuzzy rules by means of descriptors
such as form, texture and relations between objects and sub-objects. Different
approaches of classification are assessed: semantic network, selective and context
change classification. The tested site is an agricultural area near the municipality
town of Nova Esperança-Pr, with the purpose of mapping the riparian vegetation
along the Porecatú River.


Keywords


remote sensing; object oriented classification; segmentation.