A New Architecture of a Complex-Valued Convolutional Neural Network for PolSAR Image Classification
Open Access
- 1 October 2023
- journal article
- research article
- Published by MDPI AG in Remote Sensing
- Vol. 15 (19) , 4801
Abstract
Polarimetric synthetic aperture radar (PolSAR) image classification has been an important area of research due to its wide range of applications. Traditional machine learning methods were insufficient in achieving satisfactory results before the advent of deep learning. Results have significantly improved with the widespread use of deep learning in PolSAR image classification. However, the challenge of reconciling the complex-valued inputs of PolSAR images with the real-valued models of deep learning remains unsolved. Current complex-valued deep learning models treat complex numbers as two distinct real numbers, providing limited assistance in PolSAR image classification results. This paper proposes a novel, complex-valued deep learning approach for PolSAR image classification to address this issue. The approach includes amplitude-based max pooling, complex-valued nonlinear activation, and a cross-entropy loss function based on complex-valued probability. Amplitude-based max pooling reduces computational effort while preserving the most valuable complex-valued features. Complex-valued nonlinear activation maps feature into a high-dimensional complex-domain space, producing the most discriminative features. The complex-valued cross-entropy loss function computes the classification loss using the complex-valued model output and dataset labels, resulting in more accurate and robust classification results. The proposed method was applied to a shallow CNN, deep CNN, FCN, and SegNet, and its effectiveness was verified on three public datasets. The results showed that the method achieved optimal classification results on any model and dataset.Keywords
Funding Information
- National Natural Science Foundation of China (62176247, 2020-JCJQ-ZD-057-00)
- Fundamental Research Funds for the Central Universities (62176247, 2020-JCJQ-ZD-057-00)
This publication has 47 references indexed in Scilit:
- The paradigm of complex probability and Chebyshev’s inequalitySystems Science & Control Engineering, 2016
- The paradigm of complex probability and the Brownian motionSystems Science & Control Engineering, 2015
- Classification comparisons between dual-pol, compact polarimetric and quad-pol SAR imageryISPRS Journal of Photogrammetry and Remote Sensing, 2009
- An explicit fuzzy supervised classification method for multispectral remote sensing imagesIEEE Transactions on Geoscience and Remote Sensing, 2000
- A three-component scattering model for polarimetric SAR dataIEEE Transactions on Geoscience and Remote Sensing, 1998
- An entropy based classification scheme for land applications of polarimetric SARIEEE Transactions on Geoscience and Remote Sensing, 1997
- A review of target decomposition theorems in radar polarimetryIEEE Transactions on Geoscience and Remote Sensing, 1996
- Classification of multi-look polarimetric SAR imagery based on complex Wishart distributionInternational Journal of Remote Sensing, 1994
- Intensity and phase statistics of multilook polarimetric and interferometric SAR imageryIEEE Transactions on Geoscience and Remote Sensing, 1994
- Radar polaritnetry for geoscience applicationsGeocarto International, 1990