10 October 2019 in IEEE Geoscience and Remote Sensing Letters
11 January 2021 in IEEE Transactions on Aerospace and Electronic Systems
In this article, we propose a hybrid conditional random field model based on complex-valued 3-D convolutional neural network (C3NN), named as HCRF-C3NN, for polarimetric synthetic aperture radar (PolSAR) image classification. HCRF-C3NN combines the modeling power of random fields with the representation-learning ability of deep learning model for PolSAR image. By C3NN, HCRF-C3NN exploits deep amplitude and phase information of PolSAR data and, thus, measures the class probabilities in the framework of random fields. Additionally, HCRF-C3NN captures the spatial label interactions effectively in PolSAR image classification by the following work: 1) Based on the C3NN-based class probabilities, the relative entropy of class distributions is derived to describe the local structure and, thus, regulates the pairwise label interactions, which enhances the accuracy of edge location in the classification; 2) a product-of-expert potential is introduced to enforce label consistency, and, thus, HCRF-C3NN is more robust against speckle of PolSAR image and achieves smoother classification within the consistent class region. At last, to capture PolSAR image information in a more complete manner, the deep features and PolSAR scattering statistics are integrated into HCRF-C3NN based on Bayesian fusion. In this way, HCRF-C3NN effectively explores spatial correlation, scattering statistics, and deep features of PolSAR data in the framework of random fields. The experimental results demonstrate the superiority of HCRF-C3NN over the recent deep learning models for PolSAR image classification.
2 February 2021 in IEEE Transactions on Geoscience and Remote Sensing
Recently, convolutional neural network (CNN) has proved itself as a successful deep model and has been successfully utilized in polarimetric synthetic aperture radar (PolSAR) image classification. Most CNN-based models, however, concentrate on the correlation between the pixels and the labels in images and have fewer constraints on interclass or intraclass features. For fully utilizing the polarimetric data, we utilized complex-valued (CV) distance to learn the PolSAR features and proposed a PolSAR classification method by CV distance comparisons. First, we proposed a triplet CV network (TCVN) to learn the CV representations from PolSAR data by maximizing the interclass distance and minimizing intraclass distance. It uses the CV convolution and the CV Euclidean to maintain the phase components and applies the CV-dropout and CV $L_{2}$ parameter regularization to reduce the overfitting and further improve the network performance. Subsequently, CV K nearest neighbor (CV-KNN) computes the distance of the CV representations and groups similar pixels. CV-KNN is well coupled with the TCVN because both of them are based on the Euclidean distance in the complex domain. Compared with the CNN-based methods, the proposed deep metric learning model can simultaneously extract the hierarchical features by comparing the polarimetric resolution cells in the complex domain and maintain the phase component by performing CV convolutions. The effectiveness and the superiorities of CV Euclidean distance in TCVN are demonstrated. Experiments on real PolSAR images illustrate that TCVN can deal with PolSAR data more effectively and achieve comparable performance in the PolSAR image classification even with a smaller data set.
29 June 2022 in IEEE Transactions on Geoscience and Remote Sensing
Deep learning has powerful feature extraction abilities and has achieved promising results in polarimetric synthetic aperture radar (PolSAR) image classification. However, the labeled samples of PolSAR images are generally limited, which could lead to the overfitting of deep networks and the inefficiency of deep features. To overcome this problem, in this article, we propose a complex-valued enforcing population and lifetime sparsity (CV-EPLS) model to extract nonredundant sparse features from PolSAR images. CV-EPLS achieves unsupervised learning of sparse polarimetric features with limited and unlabeled samples, including amplitude and phase information in multiple polarimetric channels. Concretely, CV-EPLS defines an activation metric function to achieve strong population sparsity. Additionally, a grid search strategy is designed to ensure that activation items are evenly distributed among the sparse targets, thus forming strong lifetime sparsity. In this way, CV-EPLS constructs the complex sparse matrices and extracts discriminative sparse features in an unsupervised way, with the dependence of features being effectively reduced. Experimental results on PolSAR images demonstrate the effectiveness of CV-EPLS in the extraction of features and its application to image classification.
9 May 2022 in IEEE Geoscience and Remote Sensing Letters
For maintaining the phase information in images, complex neural networks have been widely applied to PolSAR classification. However, due to constant weights of neurons, the networks may lack randomness and be potentially overfitting for complicated imaging mechanisms and random speckle noise in PolSAR images. Thus, this letter proposes a complex variational inference network (CVIN) where complex Gaussian probability distributions are introduced into the weights of neurons in complex neural networks. In CVIN, a novel evidence lower bound (ELBO) for complex network is designed to infer the variational approximation of weights through backpropagation. After training, CVIN propagates the approximate posterior distributions given the data and makes the prediction of the labels. Thus, CVIN is an ensemble of flexible models with infinite weights, where the complex weights are regularized by the Gaussian distributions. Experiments on real PolSAR images verify the feasibility of CVIN and illustrate the potential of CVIN to serve as a competitive method for PolSAR classification.