Application of artificial neural network in the classification of coffee áreas in Machado, Minas Gerais State

• Autor(a) principal: Andrade, Livia Naiara
• Data de Publicação: 2013
• Outros Autores: Vieira, Tatiana Grossi Chquiloff, Lacerda, Wilian Soares, Volpato, Margarete Marin Lordelo, Davis Júnior, Clodoveu Augusto
• Tipo de documento: Artigo
• Idioma: eng; por
• Título da fonte: Repositório Institucional da UFLA
• Texto Completo: http://repositorio.ufla.br/jspui/handle/1/13614; http://www.coffeescience.ufla.br/index.php/Coffeescience/article/view/363
• topic: Artificial neural networks; Cafeicultura; Redes neurais artificiais; Automatic classification; Coffee; Geoprocessamento; Cafeicultura

The coffee is extremely important activity in southern of Minas Gerais and techniques for estimating acreage, seeking reliable crop forecasts are being intensely investigated. It is presented in this study, an application of Artificial Neural Networks (ANN) for the automatic classification of remote sensing data in order to identify areas of the coffee region Machado, Minas Gerais. The methodology for developing the application of RNA was divided intothree stages: pre-processing of data, training and use of RNA, and analysis of results. The first step was performed dividing the study area into two parts (one embossed busiest and least busy one with relief), because this region has a strong emphasis smooth wavy, causing a greater difficulty of automatic mapping of use earth from satelliteimages. Masks were also created in the drainage network and the urban area. In the second step, various RNA’s were trained from several samples representative of the classes of images of interest and was made to classify the rest ofthe image obtained using the best RNA. The third step consisted in analyzing and validating the results, performing across between the classified map and the map visually classified by neural network chosen. We used the Kappa index to evaluate the performance of the RNA, since the use of this coefficient is satisfactory to assess the accuracy of a thematic classification. The result was higher than the results reported in the literature, with a Kappa index of 0.558 to 0.602 relief busiest and least busy for relief.