300346520058

artigo anterior

Identificação do ecossistema mangue na Ilha do Maranhão através de técnicas de classificação utilizando imagens do sensor CCD-CBERS-2
Abstract. The mangroves are typical formations of coast under direct action of the tides, with oozy soils of estuaries areas. The need of preservation of mangrove areas has been encouraging a larger use of the remote sensing to monitor this ecosystem, through the analysis of satellite images. The main goal of this work is to evaluate different types of fraction digital processing applied into images of the sensor CCD (CBERS-2), considering the classification of the mangroves ecosystem in the Island of Maranhão-Brazil. The three original bands "RGB" composition, as well as the shadow fraction produced by Linear Spectral Mixture Model, were classified by non supervised processes (K-means and Isoseg) and supervised processes (Bhattacharyya, MaxVer and MaxVer ICM). Considering the bands, the methods and the mangroves areas we have evaluated, we have concluded the non supervised classification using the Isoseg method applied to the fraction shadow has had the best performance. This procedure has had success of 91,5% and an inclusion error of 11,2%. The classifier MaxVer-ICM applied to the shadow fraction has had a performance similar to the classifier Isoseg. However, MaxVer-ICM needs training samples, whereas Isoseg does not. Palavras-chave: Digital Classification, mangrove, CBERS-2, image processing, processamento de imagens, mangue .

1. Introdução Os manguezais são formações típicas de litoral sob ação direta das marés, com solos limosos de regiões estuárias, apresentam uma vegetação com um único estrato, de porte arbóreo e diversidade muito restrita. O mangue abriga grande variedade de espécies da fauna brasileira, como tapicuru, guará, crustáceos, sapos, insetos, garça, entre outros. O mangue garante alimento e proteção para a reprodução de inúmeras espécies marinhas e terrestres, devido ao acúmulo de