Discriminação de tipologias vegetais por meio de classificação orientada a objeto na Reserva Natural Ypeti, Paraguai

• Ano de defesa: 2014
• Autor(a) principal: Villalba Marín, Lucía Janet
• Orientador(a): Oliveira Filho, Paulo Costa de
• Tipo de documento: Dissertação
• Tipo de acesso: Acesso aberto
• Idioma: por
• Instituição de defesa: Universidade Estadual do Centro-Oeste
• Programa de Pós-Graduação: Programa de Pós-Graduação em Ciências Florestais (Mestrado)
• Departamento: Unicentro::Departamento de Ciências Florestais
• País: Brasil
• Palavras-chave em Português: GeoDMA; mineração de dados; floresta estacional semidecidual; arvore de decisão
• Palavras-chave em Inglês: GeoDMA; data minning; Atlantic Forest of Alto Paraná; decision tree
• Área do conhecimento CNPq: CIENCIAS AGRARIAS::RECURSOS FLORESTAIS E ENGENHARIA FLORESTAL; RECURSOS FLORESTAIS E ENGENHARIA FLORESTAL::MANEJO FLORESTAL
• Link de acesso: http://tede.unicentro.br:8080/jspui/handle/jspui/1358

This research aims to test the effectiveness of the technique Object–based image analysis (OBIA), using the data mining algorithms provided by GeoDMA system for mapping land cover and land use, with emphasis on the Eco Region of Upper Parana Atlantic Forest. Basically, the methodology consisted of several phases, the first was the preparation, here thes atellite image is prepared, for this research was that the sensor ALOS AVNIR 2, 4 bands, 8bit, then the processing phase was digitalized the classes identified by interpretation keys, various tests of segmentation were made, and the algorithm Baatz e Shape was chosen to test the thresholds that were in compactness 40, scale 10 and color 60, as a result 4788 polygons were obtained for the study area, after which the 51 attributes available in the GeoDMA were extracted, the classification was supervised, with the aid of the algorithm C4.5, collecting samples for each class, the classes are: semideciduous forest, riparian forest, savannah, water and roads. Subsequent collections should be calculated the attributes for the classes, and to generate the decision tree, and depending on the results of the evaluation, these activities are repeated until last to get good results. The attributes that were selected in the decision tree for better classification were: sum, ratio, mode, amplitude, dissimilarity, width and density. The Kappa index was 0,89. The GeoDMA application turns out to be very useful to do these types of tasks, for the very good accuracy in the tests, specifically in forests, and the ease interpretation by the experts, besides being a free software.