Discriminação entre carvões provenientes de madeira de Eucalyptus sp e de espécies nativas por análise de imagens digitais

• Autor(a) principal: Andrade, Bruno Geike de
• Data de Publicação: 2013
• Tipo de documento: Dissertação
• Idioma: por
• Título da fonte: LOCUS Repositório Institucional da UFV
• Texto Completo: http://locus.ufv.br/handle/123456789/3126
• topic: Classificador de bayes; Matriz de coocorrência; Análise de textura; Bayes classifier; Co-occurrence matrix; Texture analysis; CNPQ::CIENCIAS AGRARIAS::RECURSOS FLORESTAIS E ENGENHARIA FLORESTAL::TECNOLOGIA E UTILIZACAO DE PRODUTOS FLORESTAIS

The distinction between chacoal produced with Eucalyptus wood from those produced with native wood is important for the commercialization, control and supervision of charcoal production in Brazil. This work proposes the application of image analysis techniques as a new tool to assist in the separation of Eucalyptus s charcoal compared to those from other species. To conduct the study, we used 18 woody species, six belonging to the genus Eucalyptus and 12 native species in the Zona da Mata Mineira. For each species were generated 50 images of the cross section of charred wood, totaling 300 images for Eucalyptus and 600 for the native species. From these images were produced 24 images-index, combination of three different resolution (number of pixels) and eight different depths of image (number of bits per pixel). Then, 24 Bayesian classifiers for Gaussian classes were developed using texture features extracted from the images-index by the gray level co-occurrence matrix technique. The classification evaluation was performed according to the cross-validation method leaving one sample outside the training phase. Then, through the matrix of errors, were obtained Overall Accuracy and Kappa indexes that were used to evaluating the classifiers significance through a test Z. All classifiers performed better than a random classification, the 5% level of significance. The 24 classifiers developed hit, on average, 98.45% of the images tested. The use of digital image analysis, therefore, allowed to distinguish satisfactorily the chacoal produced from Eucalyptus s wood from those produced with mixed native species.