A New Entropy Estimation Method for Dual-Polarimetric SAR Data: Comparative Analysis With Quad- and Other Dual-Polarization Approaches

Abstract

This article proposes a new method for estimating target polarimetric entropy from dual-polarimetric (DP) synthetic aperture radar (SAR) data based on the reciprocity assumption commonly applied in the processing of full-polarimetric (FP) SAR data. By applying matrix analysis theory, an explicit mathematical relationship is established between the proposed DP entropy and the widely used FP entropy, based on which relationships among existing DP entropies and the FP entropy are also elucidated theoretically. Comparative experiments using FP data from San Francisco acquired by Gaofen-3 and Radarsat-2, along with DP data from the same region acquired by Sentinel-1, validate the efficacy of the theory and demonstrate the superiority of the proposed entropy in characterizing the randomness of target polarimetric scattering, which, compared to existing DP entropies, showcases better consistency with the target FP scattering characteristics and better performance for target classification in terms of both overall accuracy and reliability. Highly consistent results are also observed in the experiment based on AIRSAR data from Flevoland, which primarily focuses on analyzing the performances of DP entropies across various types of crops with low-to-medium FP entropy. Besides, this better target discrimination property of the proposed entropy also leads to its more accurate result in landslide detection, improving the capability for target feature description and disaster monitoring using DP data effectively. It is worth noting that while we apply the proposed theory to the DP data of Sentinel-1, the theoretical analysis approach is universally applicable to other DP modes as long as reciprocity holds.

Keywords

• Dual-polarimetric (DP) synthetic aperture radar (SAR)
• matrix analysis
• polarimetric scattering entropy
• reciprocity theorem
• Sentinel-1 data