SAR Image Texture Entropy Analysis for Applicability Assessment of Area-Based and Feature-Based Aea Ice Tracking Approaches

Abstract:
Precision tracking of sea ice is crucial for understanding ice dynamics and practical use in navigation through the deformation characteristics. Imaging remote sensing systems, particularly Synthetic Aperture Radar (SAR), make possible monitoring of sea ice under almost any weather conditions and independent of the illumination conditions. Two main family of approaches have been widely used so far – area-based and feature-based. We evaluate their performance using texture properties of sea ice images which are physically related to the radar signal backscattering within a SAR resolution cell. Here we use the local entropy as indicator determined by gray-level-co-occurrence matrix (GLCM) method describing how often different compositions of pixel brightness values occur in an image. A total of 120 satellite SAR sea ice images acquired over different Arctic seas from Sentinel-1 were processed and analysed. Based on the analysis, assessment criteria for choosing sea ice tracking algorithm is proposed.