An Enhancement Strategy for Wishart Classifier in Dual-Band and Dual-Pol SAR Classification

Polarimetric Synthetic Aperture Radar (PolSAR) holds significant utility in classification through exploiting polarization information. However, many sensors are dual-pol, and their capabilities of classification are limited by the absence of polarization information. This paper focuses on dual-pol classification, proposing a two-step dual-band classification strategy for the Wishart classifier. The first step involves leveraging two bands to category the strong scattering pixels. And the second step entails optimizing the classification of weak scattering pixels using interferometry coherence. We use practical data from the C and L bands, employing Support Vector Machine (SVM) as a comparative experiment to validate the effectiveness of our strategy. The total accuracy and Kappa of the Wishart classifier finally increase by 3.92% and 0.05, respectively, indicating the effectiveness of our strategy. In addition, the second step also independently improves the performance of SVM, suggesting its versatility. This study can offer a novel perspective for dual-pol and multi-band classification.