Open Journal Systems

USO INTEGRADO DE IMAGEM DE ALTA RESOLUÇÃO ESPACIAL E ALTURA DERIVADA DO LASER SCANNER NA ESCOLHA DO CLASSIFICADOR ORIENTADO A REGIÃO

MOSAR FARIA BOTELHO, JORGE ANTONIO SILVA CENTENO

Abstract



Neste artigo são apresentados os resultados de um estudo que aborda o tema da
integração das alturas derivadas de um levantamento laser scanner com imagens de
satélite de alta resolução espacial, Quickbird II, para o mapeamento temático de
áreas urbanas. Justifica-se este trabalho, devido o uso de imagens orbitais não ser
suficiente para distinguir os objetos associados ao mapeamento em áreas urbanas,
de forma semi-automática, tornando-se necessária a inclusão de outras informações
auxiliares. Para isto, uma metodologia orientada à análise de regiões na imagem é
proposta. O ganho obtido com a inclusão da altura derivada do laser scanner é
comparado com o ganho resultante do uso de parâmetros espaciais derivados das
imagens, mediante a utilização do algoritmo de classificação da rede neural artificial
(RNA) e classificadores convencionais (máxima verossimilhança, distância
mínima). Os resultados mostram que, a contribuição da informação altimétrica é
muito útil na classificação de edificações, vias e vegetação. Dentre os algoritmos
testados, o uso de RNA incluindo informações de elevações derivadas do laser
scanner foi o que forneceu os melhores resultados.

Integrated Use Of Spatial High Resolution Image And Laser Scanner Derived
Hight Classificator Oriented To Region To Choose

Abstract


In this article, the results of a research about the integration of height derived from a
laser scanner measurement with high resolution spatial sattelite image are
presented, QUICKBIRD II, for the tematic mapping of urban areas. This research is
justified, for so as not to suffice the use of orbital images, to distinguish the associated objects to the urban area mapping, in a semi-automatic way, one needs
the inclusion of other auxiliary information. Therefore, it is proposed a metodology
oriented to a regional analysis in the image. The gain obtained with the laser
scanner derived height inclusion is compared to the resulting gain obtained from the
use of spatial parameters derived from the images, through the use an artificial
neural network classification algorithm (ANN) and usual classifiers (maximum
likelihood and minimum distance). The results show that the altimetry information
contribution is very useful in building, road and vegetation classification. From the
tested algorithms, the use of an ANN including information of height derived from
a laser scanner was the one which got the best results.