Multi-View Feature Selection for PolSAR Image Classification via l₂,₁ Sparsity Regularization and Manifold...

Area of Science:

• Remote Sensing
• Computer Vision
• Machine Learning

Background:

• Polarimetric synthetic aperture radar (PolSAR) image classification relies heavily on feature extraction.
• High-dimensional data from polarimetric features (PF) and texture features (TF) lead to computational complexity and reduced classification performance.
• Effective feature selection is crucial for improving PolSAR image classification accuracy.

Purpose of the Study:

• To propose a multi-view feature selection method for PolSAR image classification.
• To address the challenges of high dimensionality and feature redundancy in PolSAR data.
• To enhance the efficiency and accuracy of PolSAR image classification.

Main Methods:

• Generation of two distinct feature types: polarimetric features (PF) and texture features (TF).
• Development of an optimization model to identify optimal feature selection matrices.
• Implementation of l_2,1 norm sparsity regularization for feature selection and manifold regularization to preserve data structure.

Main Results:

• The proposed multi-view feature selection method was evaluated on three real PolSAR datasets.
• Experimental results demonstrated the superiority of the proposed method in PolSAR image classification.
• The method effectively selects relevant features, reducing computational complexity and improving classification performance.

Conclusions:

• The developed multi-view feature selection approach offers a significant advancement in PolSAR image classification.
• The method's ability to handle high-dimensional data and preserve structural information leads to superior classification outcomes.
• This technique provides a robust solution for enhancing the analysis of PolSAR imagery.