A Classification of Polarimetric SAR Images Based on Cloude Decomposition and Domain-Adversarial Neural Network
ZHAO Mingjun
ZHANG Linlin
LI Shen
LI Qin
XIAO Qiang
Abstract：Traditional supervised classification method is usually used to an assumption that the probability distributions of training data and testing data are consistent.Such a method is significant limited,whereas it is caught in the domain discrepancy problems caused by heterogeneous polarimetric synthetic aperture radar(SAR)images.For the mentioned-above reasons,this paper innovatively proposes a polarimetric SAR image classification method in combination of Cloude decomposition with a deep domain-adversarial neural network(DANN).Firstly,the method is to utilize the features obtained from Cloude decomposi-tion for taking as an input of the DANN.Through the adversarial training of the domain classifier and the class classifier in the DANN,domain-invariant features are extracted,and then the classification of hetero-geneous polarimetric SAR images is achieved.The experimental results show that the method effectively overcomes the limitations of the traditional supervised method in domain adaptation.In two sets of trans-fer comparison experiments involving five actual polarimetric SAR images,the overall classification accu-racy is significantly enhanced compared to the direct transfer method and the DANN method utilized raw data as input.
Keywords：polarimetric SAR image classificationdomain-adversarial neural networktransfer learning
Publication Date：2025-12-25
Online Publishing Date：2025-12-26(First online date of this platform, not the publication date of the document)
Pages：10( 65-74 )