Classificação de áreas irrigadas por pivôs centrais utilizando como base a segmentação

Abstract

The use of techniques of analysis of remote sensing data based on homogeneous regions segmentation opens new perspectives for center pivots irrigated areas estimation, because of their well-defined shapes. The goal of this work was to compare and assess maps of irrigated areas by center pivots using two interpretation methods: visual interpretation and a classification based on segmentation techniques. The thematic maps were generated from CCD/CBERS-2B satellite images. The steps of the work involved the geo-reference and the image geometric correction, the visual interpretation of irrigated areas, segmentation, and supervised classification for extraction of features. The irrigated areas were presented in different colors and shades, representing agricultural crops in different stages and soil exposed. The total values of irrigated areas were 631.38 ha, for the manual method and 669.12 ha for the segmentation method. It was concluded that there was an excellent agreement between the areas that were delineated manually and the areas extracted with the method based on segmentation, with kappa index of 94%. The method based on segmentation did not separate the center pivots, but allowed to identify and estimate the total irrigated areas. The extraction of irrigated areas by center pivots, using the process of segmentation, depended on the spatial distribution of the objects.