Classificador de máxima verossimilhança aplicado à identificação de espécies nativas na Floresta Amazônica.

• Autor(a) principal: MORAS FILHO, L. O.
• Data de Publicação: 2017
• Outros Autores: FIGUEIREDO, E. O., ISAAC JÚNIOR, M. A., BARROS, V. C. C. de, HOTT, M. C., BORGES, L. A. C.
• Idioma: por
• Título da fonte: Repositório Institucional da EMBRAPA (Repository Open Access to Scientific Information from EMBRAPA - Alice)
• Texto Completo: http://www.alice.cnptia.embrapa.br/alice/handle/doc/1072220
• topic: Manejo florestal; Método de classificação digital; Maximum Likelihood; Máxima verossimilhança; Rio Branco (AC); Acre; Amazônia Ocidental; Western Amazon; Amazonia Occidental; Sistemas de información geográfica; Identificación de plantas; Estimación; Especies nativas; Bosques tropicales; Análisis estadístico; Teledetección; Floresta tropical; Espécie nativa; Identificação; Estimativa; Sensoriamento remoto; Sistema de informação geográfica; Análise estatística; Método estatístico; Tropical forests; Indigenous species; Plant identification; Estimation; Remote sensing; Geographic information systems; Statistical analysis

Among a variety of digital classification methods based on remote sensing images, the Maximum Likelihood (ML) is widely used in environmental studies, mainly for land cover and vegetation analysis. This study aimed to evaluate the effectiveness of supervised classification by ML technique in a forest management area of dense ombrophilous forest, using one RapidEye image. With this purpose, it was conducted the census of species over 30 cm in diameter at breast height and calculated the Cover Value Index (CVI), and selected the 20 species with the highest CVI as a parameter for classification in a Geographic Information System. 13 of the 20 species selected in the study area were not identified by the classification method, and among the seven identified species, two were underestimated and the others were overestimated. Both the maximum likelihood technique and the spatial resolution of the image used were not suitable for supervised classification of native vegetation, with Kappa index of 0.05 and global accuracy of 5.53%. Studies using spectral characterization in leaf level supported by higher or hyper spectral and spatial resolution images are recommended to increase the accuracy of classification.