Revista Brasileira de Cartografia

Water management is currently a key field to support life and economic activity. The increased mechanization in agriculture, mainly through center pivot irrigation systems, it is a challenge, because consume the majority of hydrological resources. The Brazil is among the 10 largest in terms of irrigated area (8.2 × 10 ha), making monitoring essential. Thus, we present an approach based on digital image processing and machine learning for detecting center pivots. The methodology focuses on circular Hough transform (CHT) and balanced random forest (BRF) classifier using vegetation indices (NDVI/SAVI) generated from Landsat 8 images and land use and land cover (LULC) data provided by MapBiomas project . The candidate's pivots circles identified on images are filtered based on the spectral response of the vegetation and the shape characteristics of objects present in these areas. Our approach was able to detect 7358 pivots, reaching an 83.86% recall for the 52 tiles analyzed in Brazil compared with mapping performed by the Brazilian National Water and Sanitation Agency (ANA). In some tiles, the recall reached up to 100%. The BRF model trained with spectral and geometric features allowed for the identification of pivots, where regions with a great amplitude of vegetation indices highlight areas with agricultural activity to the detriment of areas of native vegetation, and characteristics of the shapes from targets based on their delimitation through the High Pass Filter Sharr. The good accuracy achieved shows the robustness of the method in detecting pivots on a large spatial and temporal scale.