Deep Mutual Information Subspace Clustering Network for Hyperspectral Images

Graph Regularized Residual Subspace Clustering Network for hyperspectral image clustering

There is a relative paucity of studies that seek to nonparametrically describe or categorize hyperspectral image (HSI) data in the most natural context within which those data are best represented: n-dimensional (n-D) hyperspace, where n is the number of bands in an HSI data cube. Statistical techniques-and specifically those based on second-order statistics-are predominant and generally do not provide (nor are they utilized to provide or 'leave behind') an understanding of the distribution of spectra in hyperspace for the purpose of discovering patterns, comparing data sets, developing a deeper understanding of HSI data, etc. With the terabytes of HSI data in archives today representing a huge diversity of geographical settings, it seems natural to enquire as to whether there are universal patterns among HSI data sets. If such patterns exist, new techniques may be developed to exploit them, e.g., to enhance information extraction speed and accuracy. Since an understanding of the nature of HSI data has rarely gone beyond cursory, visual assessments of geographical setting; i.e., assessments of whether the data are of desert, forest, urban, littoral, etc., environments, it is perhaps worthwhile to develop methods to probe n-D hyperspace with the intent to identify and describe universal data commonalities, differences, classifications, and patterns. Building upon earlier work (Resmini, 2003 [1], 2006 [2]), an algorithm was created to describe the volume (and label the individual volume elements) of hyperspace occupied by an HSI data set. The algorithm essentially discretizes hyperspace assigning addresses to boxes in which data points (i.e., spectra) reside. The algorithm has been applied to visible/near-infrared to shortwave infrared (VNIR/SWIR) and longwave infrared (LWIR) HSI data cubes. Implemented in C code, the algorithm may also be used to calculate the fractal dimension of any array of points in any n-D hyperspace because it is essentially an implementation of box counting. The algorithm is described as are results of its application to actual HSI data sets. The fractal dimension of HSI data derived from the algorithm are also presented and compared to principal componentsbased estimates of data dimensionality. Application of the algorithm for developing a deeper understanding of the nature of HSI data as viewed as points in hyperspace is discussed as are practical applications for 'traditional' HSI exploitation activities such as anomaly and target detection.

Deep subspace clustering methods have demonstrated its outstanding capability for hyperspectral image (HSI) clustering. However, there is a lack of explicit supervision to ensure that low-dimensional spatial–spectral features learned from high-dimensional HSI cubes have good subspace structures. In this letter, we propose a cascade residual capsule network (CRCN) for extracting deep spatial–spectral features and introduce the coding rate reduction (CRR) to measure the compactness of learned spatial–spectral features. We exploit the difference between the coding rate of all features and the sum of that of features of each category as loss function, which provides explicit supervision for learning invariant spatial–spectral features to HSI cube flips and rotations and facilitate the subsequent HSI clustering. To learn features that are discriminative to diverse spatial context information on land-cover objects of the same category, we add a regularization term to the loss function, which is a brink loss between the <inline-formula xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink"> <tex-math notation="LaTeX">${l_{2}}$ </tex-math></inline-formula> -norm of active vectors of class capsules in the proposed CRCN and the assigned labels. Experimental results on three benchmark HSI datasets demonstrate the effectiveness of the proposed method, which achieves superiority performance over several state-of-the-art HSI clustering methods.

Hyperspectral image (HSI) clustering has been widely used in the field of remote sensing. However, most traditional clustering algorithms are not suitable for dealing with large-scale HSI due to their low clustering performance and high computational complexity. In this letter, we propose a novel fuzzy clustering algorithm, called fuzzy embedded clustering based on the bipartite graph (FECBG), to efficiently deal with large-scale HSI clustering problems. First, we propose the FECBG method that incorporates fuzzy clustering with the nonnegative regularization term based on the bipartite graph into a unified model, which has good clustering performance and reduces the sensitivity of fuzzy clustering to the initial cluster centers. Second, we adopt the fast spectral embedded method to obtain the low-dimensional representation of HSI data, to reduce the computational complexity. At last, we add the nonnegative regularization term based on the bipartite graph to fuzzy clustering, to constrain the solution space of the fuzzy membership matrix. Experimental results on several HSI data sets have demonstrated the efficiency and effectiveness of the proposed FECBG algorithm.

Hyperspectral image clustering is usually used for unsupervised learning in different applications. However, the traditional clustering methods have not been considered the complex relationships among neighbouring pixels. The Albedo and Shading elements can define pixel values in the HyperSpectral Images (HSIs). In HSIs, features are different from each other because of their natural physical characteristics and the physical nature of different image features can be described by the Albedo element. Therefore, in this paper, we generate the natural Albedo feature of the HSIs by applying Albedo recovery step to exploit main information from HSIs. Then, we utilized the Fuzzy C-means clustering method to cluster the natural Albedo dataset. In this paper, we propose a novel accurate Albedo Recovery based Fuzzy C-Means (ARFCM) method to cluster HSIs. In the dataset, each feature vector is processed by the Albedo recovery step to create a new feature vector. This new feature vector can describe the dataset better than the original one. Comparing clustering methods as one of the powerful clustering algorithms are widely used in the remote sensing fields of studying. The experiments conducted on several benchmark datasets demonstrated that the proposed clustering method achieves higher performance than other methods and present the efficiency and effectiveness of the proposed method. The results of experiments over different HSI datasets indicated that the proposed method could produce reliable and suitable results compared to the other methods. This shows the robustness of the proposed ARFCM algorithm over the various HSI datasets. Other methods may provide a good response in a given dataset and do not perform well in the other data. Consequently, the ARFCM method, regardless of the study area characteristics and the sensor features, always renders remarkable clustering accuracy.

Hyperspectral image (HSI) clustering is a challenging task due to the complex spatial-spectral structure and high-dimensional property in HSI data. In this letter, a novel clustering information-constrained 3D U-Net subspace clustering network is proposed for HSI clustering. Considering the spatial-spectral information, the proposed network takes the 3D pixel cubes around the pixels as the input. Based on the 3D pixel cubes, a 3D U-Net subspace clustering network is introduced to extract spatial-spectral features from 3D pixel cubes and learn self-representation subspace property among pixels. In order to learn features more suitable for clustering, a clustering information constraint is introduced to explore useful information gain in the existing clustering result. Experiments conducted on three public HSI datasets illustrate the superior performance of the proposed method.

Coupled adversarial learning for fusion classification of hyperspectral and LiDAR data

Deep neural networks (DNNs) have revolutionized the way remotely sensed hyperspectral image (HSI) data are managed and processed. For instance, residual networks (ResNets) have achieved high classification accuracy by applying sequential transformations (layer by layer) on the input HSI data, obtaining highly discriminative data representations. However, these models are quite complex, with significant requirements in terms of memory resulting from the large number of parameters that they need to learn, which also leads to potential overfitting issues. In this work, we specifically address the aforementioned problem by re-interpreting a DNN (the ResNet) as a continuous transformation, instead of the traditional (discrete) step-by-step approach. To achieve this, we combine ordinary differential equations (ODEs) with DNN architectures for the first time in the HSI data classification literature. This allows us to perform remotely sensed HSI data classification in an efficient way in terms of number of parameters. Our experimental results, conducted using two well-known HSI data sets, indicate that the inclusion of ODEs in the architecture of DNNs offers significant advantages when processing and classifying this kind of high-dimensional data, achieving better performance even with less training data.

Hyperspectral image (HSI) clustering groups pixels into clusters without labeled data, which is an important yet challenging task. For large-scale HSIs, most methods rely on superpixel segmentation and perform superpixel-level clustering based on graph neural networks (GNNs). However, existing GNNs cannot fully exploit the spectral information of the input HSI, and the inaccurate superpixel topological graph may lead to the confusion of different class semantics during information aggregation. To address these challenges, we first propose a structural-spectral graph convolutional operator (SSGCO) tailored for graph-structured HSI superpixels to improve their representation quality through the co-extraction of spatial and spectral features. Second, we propose an evidence-guided adaptive edge learning (EGAEL) module that adaptively predicts and refines edge weights in the superpixel topological graph. We integrate the proposed method into a contrastive learning framework to achieve clustering, where representation learning and clustering are simultaneously conducted. Experiments demonstrate that the proposed method improves clustering accuracy by 2.61%, 6.06%, 4.96% and 3.15% over the best compared methods on four HSI datasets. Our code is available at https://github.com/jhqi/SSGCO-EGAEL.

This paper presents a novel clustering method, named Laplacian regularized deep subspace clustering (LRDSC), for unsupervised hyperspectral image (HSI) classification. We introduce the Laplacian regularization into the subspace clustering to consider the manifold structure reflecting geometric information. To enable the subspace clustering, which works in linear space, to deal with the complicated HSI data with nonlinear characteristics, we combine the subspace clustering as a self-expressive layer with deep convolutional auto-encoder. Furthermore, the 3-D convolutions and deconvolutions with skip connections are utilized to make full extraction of the spectral-spatial information and full use of the historical feature maps produced by the network. We compare the results of the proposed method with six existing cluster methods on four real hyperspectral data sets, showing that the proposed method is able to achieve state-of-the-art performance.

Hyperspectral image (HSI) data have large numbers of bands that probably not all bands are equally informative and predictive for an effective HSI classification. Effective algorithms are highly desired in many real-world HSI applications, especially in cases requiring rapid learning with limited computing power. To address the abovementioned case, we present in this article a novel plug-and-play compact band weighting (CBW) module based on the attention-driven mechanism that evaluates different spectral bands according to their contributions to a given classification task. Compared to existing band weighting (BW) modules with tens of thousands of network parameters by deep learning, the proposed CBW is a lightweight module with only 20 parameters. Both model complexity and time cost are significantly reduced. The CBW module implements BW by making full use of the correlation among the adjacent spectral bands and spectral statistic information and, thereby, leads to the effect of recalibrated HSI. The experimental study has been conducted on three widely used HSI data sets, and results show the superiority of the proposed algorithm over current state-of-the-art methods of BW. The source code is available at <uri xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">https://github.com/JarvenYi/CBW</uri> .

Many food products are subjected to quality control analyses for detection of surface residue/contaminants, and there is a trend of requiring more and more documentation and reporting by farmers regarding their use of pesticides. Recent outbreaks of food borne illnesses have been a major contributor to this trend. With a growing need for food safety measures and “smart applications” of insecticides, it is important to develop methods for rapid and accurate assessments of surface residues on food and feed items. As a model system, we investigated detection of a miticide applied to maize leaves and its miticidal bioactivity over time, and we compared two types of reflectance data: fourier transformed infrared (FTIR) data and hyperspectral imaging (HI) data. The miticide (bifenazate) was applied at a commercial field rate to maize leaves in the field, with or without application of a surfactant, and with or without application of a simulated “rain event”. In addition, we collected FTIR and HI from untreated control leaves (total of five treatments). Maize leaf data were collected at seven time intervals from 0 to 48 hours after application. FTIR data were analyzed using conventional analysis of variance of miticide-specific vibration peaks. Two unique FTIR vibration peaks were associated with miticide application (1,700 cm−1 and 763 cm−1). The integrated intensities of these two peaks, miticide application, surfactant, rain event, time between miticide application, and rain event were used as explanatory variables in a linear multi-regression fit to spider mite mortality. The same linear multi-regression approach was applied to variogram parameters derived from HI data in five selected spectral bands (664, 683, 706, 740, and 747 nm). For each spectral band, we conducted a spatial structure analysis, and the three standard variogram parameters (“sill”, “range”, and “nugget”) were examined as possible “indicators” of miticide bioactivity. We demonstrated that both FTIR peaks and standard variogram parameters could be used to accurately predict spider mite mortality, but linear multi-regression fits based on standard variogram parameters had the highest accuracy and were successfully validated with independent data. Based on experimental manipulation of HI data, the use of spatial structure analysis in classification of HI data was discussed.

Traditional clustering methods groups data points according to attributes such as similarity, continuity, neighbor-hood information, etc. overlooks the structural properties of the data. Consequently, prevalent clustering approaches to below-par performance in real-world applications. Unlike traditional clustering approaches, subspace clustering methods attempt to group datapoints keeping the inherent structure and rank-related properties of the data into account. Despite the rapid growth in deep learning-based approaches, very few works have utilized deep learning for the subspace clustering task. This work introduced an auto-encoder-based deep learning architecture consisting of a self-expressive layer for the deep subspace clustering task. We use smoothed L <inf xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">2</inf> , L0.5 and Frobenius norms instead of the actual measures for ease of optimization task. We also explored the efficacy of sparsity measures that characterize the self-representation coefficient matrix of the self-expressive layer. The experiments conducted on standard datasets suggest that the application of efficient sparsity measures improves the performance of the subspace clustering approach and results in superior performance compared to the previous deep subspace clustering approaches.

Hyperspectral remote sensing organically combines traditional space imaging with advanced spectral measurement technologies, delivering advantages stemming from continuous spectrum data and rich spatial information. This development of hyperspectral technology takes remote sensing into a brand-new phase, making the technology widely applicable in various fields. Hyperspectral clustering analysis is widely utilized in hyperspectral image (HSI) interpretation and information extraction, which can reveal the natural partition pattern of pixels in an unsupervised way. In this article, current hyperspectral clustering algorithms are systematically reviewed and summarized in nine main categories: centroid-based, density-based, probability-based, bionics-based, intelligent computing-based, graph-based, subspace clustering, deep learning-based, and hybrid mechanism-based. The performance of several popular hyperspectral clustering methods is demonstrated on two widely used data sets. HSI clustering challenges and possible future research lines are identified.

Subspace clustering has become widely adopted for the unsupervised analysis of hyperspectral images (HSIs). Recent model-aware deep subspace clustering methods often use a two-stage framework, involving the calculation of a self-representation matrix with complexity of O(n2), followed by spectral clustering. However, these methods are computationally intensive, generally incorporating only local or non-local structure constraints, and their structural constraints fall short of effectively supervising the entire clustering process. We propose a scalable, context-preserving deep clustering method based on basis representation, which jointly captures local and non-local structures for efficient HSI clustering. To preserve local structure—i.e., spatial continuity within subspaces—we introduce a spatial smoothness constraint that aligns clustering predictions with their spatially filtered versions. For non-local structure—i.e., spectral continuity—we employ a mini-cluster-based scheme that refines predictions at the group level, encouraging spectrally similar pixels to belong to the same subspace. These two constraints are jointly optimized to reinforce each other. Specifically, our model is designed as a one-stage approach, in which the structural constraints are applied to the entire clustering process. The time and space complexity of our method are O(n), making it applicable to large-scale HSI data. Experiments on real-world datasets show that our method outperforms state-of-the-art techniques.