Simulated multipolarized MAPSAR images to distinguish agricultural crops

• Autor(a) principal: Silva, Wagner Fernando
• Data de Publicação: 2012
• Outros Autores: Rudorff, Bernardo Friedrich Theodor, Formaggio, Antonio Roberto, Paradella, Waldir Renato, Mura, José Claudio
• Tipo de documento: Artigo
• Idioma: eng
• Título da fonte: Scientia Agrícola (Online)
• Texto Completo: https://www.revistas.usp.br/sa/article/view/22767
• topic: SAR; crop discrimination; multipolarized L-band; radar; remote sensing

Many researchers have shown the potential of Synthetic Aperture Radar (SAR) images for agricultural applications, particularly for monitoring regions with limitations in terms of acquiring cloud free optical images. Recently, Brazil and Germany began a feasibility study on the construction of an orbital L-band SAR sensor referred to as MAPSAR (Multi-Application Purpose SAR). This sensor provides L-band images in three spatial resolutions and polarimetric, interferometric and stereoscopic capabilities. Thus, studies are needed to evaluate the potential of future MAPSAR images. The objective of this study was to evaluate multipolarized MAPSAR images simulated by the airborne SAR-R99B sensor to distinguish coffee, cotton and pasture fields in Brazil. Discrimination among crops was evaluated through graphical and cluster analysis of mean backscatter values, considering single, dual and triple polarizations. Planting row direction of coffee influenced the backscatter and was divided into two classes: parallel and perpendicular to the sensor look direction. Single polarizations had poor ability to discriminate the crops. The overall accuracies were less than 59 %, but the understanding of the microwave interaction with the crops could be explored. Combinations of two polarizations could differentiate various fields of crops, highlighting the combination VV-HV that reached 78 % overall accuracy. The use of three polarizations resulted in 85.4 % overall accuracy, indicating that the classes pasture and parallel coffee were fully discriminated from the other classes. These results confirmed the potential of multipolarized MAPSAR images to distinguish the studied crops and showed considerable improvement in the accuracy of the results when the number of polarizations was increased.