Comparative analysis of polarimetric SAR images based on multi-target decomposition

Document Type : Original Article

Authors

1 Electrical Engineering Department, Military Technical College, Cairo, Egypt.

2 National Authority of Remote Sensing and Space Sciences, 23 Joseph Tito St, Cairo, Egypt.

10.1088/asat.2023.344357

Abstract

A novel approach for PolSAR image analysis using support vector machines (SVM) were presented in this paper, with a focus on the impact of target decomposition techniques on classification accuracy. We explore the use of six different target decomposition techniques, including Cloude, Huynen, HAAlpha, Freeman, Vanzyl, and Yamaguchi, to extract feature vectors for training SVM models. Our study evaluates the performance of the classifiers on two standard benchmark datasets (Flevoland and San Francisco Bay) using multiple assessment metrics, including accuracy, sensitivity/recall, specificity, precision, F1-score, and Kappa coefficient. Our contribution is twofold: first, we provide a comprehensive analysis of how the choice of target decomposition technique affects the classification accuracy of PolSAR
images using SVMs, and second, we demonstrate the effectiveness of SVMs for PolSAR image classification, particularly for differentiating between different land cover types. Our results show that certain target decomposition techniques are better suited for specific land cover types, and our approach can achieve high classification accuracy across different datasets. Overall, our study provides important insights into the effective use of SVMs and target decomposition
techniques for PolSAR image analysis.

Keywords

Main Subjects

International Conference on Aerospace Sciences and Aviation Technology
Volume 20, Issue 20
20th International Conference on Aerospace Sciences & Aviation Technology (ASAT-20). "Special Issue"
March 2023
Pages 1-16

Files

Share

How to cite

Statistics