Entropy/anisotropy/alpha based 3DGabor filter bank for PolSAR image classification

Abstract

There are two types of features in a polarimetric synthetic aperture radar (PolSAR) image: 1-physical scattering characteristics of radar, and 2-geometric and texture properties. To handle the 3D nature of PolSAR image, a revised 3DGabor filter bank is proposed here. Although the 3DGabor filter can explore the interaction between three dimensions of the PolSAR cube, but it does not utilize the physical information of radar including the scattering mechanism. To deal with this issue, the physical parameters of the entropy/anisotropy/alpha decomposition theory are used to revise the 3DGabor filter. The output is joint texture and physical features of the PolSAR cube. These features are highly discriminative such that they can provide high accurate classification maps specially by using limited training samples. The experimental results on three PolSAR images show better performance of the proposed method with respect to several competitors with a significant difference from the statistical point of view.

Keywords:

• Entropy/anisotropy/alpha
• 3DGabor filter
• classification
• PolSAR

Data availability statement

No new datasets were generated or analyzed in this article. The datasets used for the experiments are publicly available benchmark datasets.

Disclosure statement

No potential conflict of interest was reported by the authors.