Estimação do volume de madeira empilhada utilizando imagens digitais e redes neurais

• Ano de defesa: 2014
• Autor(a) principal: Silveira, Daniel de Paula
• Orientador(a): Silva, Márcio Lopes da
• Banca de defesa: Leite, Hélio Garcia , Binoti, Daniel Henrique Breda
• Tipo de documento: Dissertação
• Tipo de acesso: Acesso aberto
• Idioma: por
• Instituição de defesa: Universidade Federal de Viçosa
• Programa de Pós-Graduação: Mestrado em Ciência Florestal
• Departamento: Manejo Florestal; Meio Ambiente e Conservação da Natureza; Silvicultura; Tecnologia e Utilização de
• País: BR
• Palavras-chave em Português: Eucalipto - Madeira; Madeira - Fator empilhamento; Empilhamento - Volume sólido; Processamento imagens - Técnicas digitais; NeuroDIC (Programa de computador)
• Palavras-chave em Inglês: Eucalyptus - Madeira; Wood - stacking factor; Stacking - Solid volume; Processing images - Digital Techniques; NeuroDIC (computer program)
• Área do conhecimento CNPq: CNPQ::CIENCIAS AGRARIAS::RECURSOS FLORESTAIS E ENGENHARIA FLORESTAL::MANEJO FLORESTAL
• Link de acesso: http://locus.ufv.br/handle/123456789/3169

In this study is presented the computerized system to estimate the piling factor using artificial neural networks - NeuroDIC and use it to estimate the volume of wood piled on trucks using digital images. NeuroDIC was developed using the Java programming language. As a development environment IDE (Integrated Development Environment) Netbeans 7.1 and JDK 7.3 (Java Development Kit) was used. The software is a free tool that can be used on any operating system. The wood used in the study came from a field belonging to a reforestation company Midwestern São Paulo planted with a hybrid of Eucalyptus grandis x Eucalyptus urophylla, aged 8.5 years, planting space of 3.0 X 3.0 meters, piled on trucks. Each stack loaded on the trucks was photographed across the length of the logs, a distance of approximately 2.5 meters. In each pile, six photographs were taken, one from the upper left (SE) , upper center (SC), upper right (SD), lower left (IE), lower right (ID) and a center (CE). The cubic volume of each pile was estimated by cubing each log by the method of Huber. The stereo volume was calculated by measuring the dimensions of the pile in the truck. Altogether 420 networks were trained and chosen 42 among them, one for each image. The networks have been applied and each classified image was visually compared to the original image, analyzing the quality of the classification after application of neural networks. 42 piling factors were obtained in total. The average between them was calculated resulting in medium piling factor. The average difference between the calculated cubic volume in the field and by digital image processing using RNA was 1.9%. Compared to the conventional process of measuring the entire load to determine the piling factor, there was a reduction of approximately 90% of activity costs to adopt the digital image processing using artificial neural networks.