Research and simulation of images classification algorithm for distributed objects obtained by remote sensing of the Earth's surface

Abstract

Application of ultra-wideband signals and large apertures makes it possible to obtain a sufficiently detailed radar image of a spatially distributed object at the processing unit output. But the problem of optimal classification of synthetic aperture radar (SAR) data remains relevant due to specific features of radar images. The paper describes solution of the synthesis problem for optimal SAR image classification algorithm. The optimal set of very informative and small dimension features is found and features based on the moments of SAR images are proposed. A comparative analysis of classification algorithms using various features is made, and the ratio is proposed, which can be used as the classification sign that is invariant to object shift and to specific distortions caused by the object rotation. © 2016 Academic Publications, Ltd.