FEASIBILITY OF EMPLOYING ARTIFICIAL NEURAL NETWORKS FOR EMERGENT CROPMONITORING IN SAR SYSTEMS

An investigation into the feasibility of using high-resolution synthet ic aperture radar (SAR) data and artificial neural networks for monito ring the stage of growth of a crop is presented. The high resolution d ata sets representing an experimentally simulated crop at three differ ent stages of growth are acquired at X-band by means of a ground-based synthetic aperture radar (GB-SAR) system under development at the Uni versity of Sheffield. A hybrid classification system, developed in rec ent studies, is then applied to these image sets, providing very high training and test data accuracy (95.8% and 94.4%, respectively) for di fferences in growth of the order of a quarter of a wavelength, and acc eptable results (79.9% and 71.9%, respectively) for differences of the order of a tenth of a wavelength. The procedures developed for the hi gh-resolution data acquisition are described and the results obtained by applying the hybrid classification system to the acquired data are discussed.