INFLUENCE OF BLOCK DIMENSION ON THE CLASSIFICATION OF DIGITAL IMAGES

• Autor(a) principal: Reis, Leonardo Rubim
• Data de Publicação: 2009
• Outros Autores: Vieira, Luciano Baião, Schaefer, Carlos Ernesto, Santos, Nerilson Terra, Pinto, Francisco Assis
• Tipo de documento: Artigo
• Idioma: por
• Título da fonte: Engenharia na Agricultura
• Texto Completo: https://periodicos.ufv.br/reveng/article/view/59
• topic: Mecanização Agrícola

To determine the effect of block dimensions on image classification, aerial images from three different heights (4, 20 and 50 m) were obtained in an experimental area of 1.200 m2, divided into three blocks, each with five soil preparation methods. The images were obtained with the use of a digital camera in a balloon. The block dimensions for heights of 4 and 20 m were respectively, 50 x 50, 150 x 150 and 250 x 250 pixels and for the height of 50 m, 50 x 50 and 90 x 90 pixels. The images were processed with the use of co-occurrence matrix techniques from which eight texture descriptors, calculated from the image blocks, of the images were extracted for classification. The highest values of the kappa index for 4 and 20 m height were obtained from block of 250 x 250 pixels. For the height of 50 m, the 90 x 90 block gave the best result.