Band Selection Of Hyperspectral Images Research Articles

Enhanced Wild Horse Optimizer with Cauchy Mutation and Dynamic Random Search for Hyperspectral Image Band Selection

The high dimensionality of hyperspectral images (HSIs) brings significant redundancy to data processing. Band selection (BS) is one of the most commonly used dimensionality reduction (DR) techniques, which eliminates redundant information between bands while retaining a subset of bands with a high information content and low noise. The wild horse optimizer (WHO) is a novel metaheuristic algorithm widely used for its efficient search performance, yet it tends to become trapped in local optima during later iterations. To address these issues, an enhanced wild horse optimizer (IBSWHO) is proposed for HSI band selection in this paper. IBSWHO utilizes Sobol sequences to initialize the population, thereby increasing population diversity. It incorporates Cauchy mutation to perturb the population with a certain probability, enhancing the global search capability and avoiding local optima. Additionally, dynamic random search techniques are introduced to improve the algorithm search efficiency and expand the search space. The convergence of IBSWHO is verified on commonly used nonlinear test functions and compared with state-of-the-art optimization algorithms. Finally, experiments on three classic HSI datasets are conducted for HSI classification. The experimental results demonstrate that the band subset selected by IBSWHO achieves the best classification accuracy compared to conventional and state-of-the-art band selection methods, confirming the superiority of the proposed BS method.

Integration of Spatial Spectral Structure with Improved Local Density for Hyperspectral Image Band Selection Method

Integration of Spatial Spectral Structure with Improved Local Density for Hyperspectral Image Band Selection Method

Hyperspectral image band selection method combining the variation degree of salient features

ABSTRACTBand selection (BS) work for hyperspectral remote sensing images (HRSIs) has attracted increasing attention from researchers. In the HRSI, the salient features of each band always include the more important discriminative features. However, most existing BS methods always have difficulty in simultaneously performing local-global consistency analysis on salient features in bands of HRSIs. In addition, these methods also ignore the consideration of the diversity of band information to some extent when selecting a band subset. To tackle these problems, a novel unsupervised hyperspectral BS method combining the variation degree of salient features (UBSF) is proposed in this article. UBSF first promotes the singular value decomposition method to analyse the salient features in bands and develops a new low-dimensional local-global salient feature representation (NLR) method for discriminative feature representations of the bands. The NLR method combines the variation degree of the features of the significant representative patches of each band in their corresponding significant feature subspaces to achieve a local-global consistency analysis of salient features of each band of HRSIs. Next, UBSF proposes a new two-stage strategy (TSS) to preserve the diversity of BS results as much as possible. In TSS, we propose a novel similarity criterion, which can make the highly similar band subsets with near information amount in the overall bands adaptively compressed according to the obtained similarity results of bands further. Then, the band subset with the most information is selected as the final BS results in the obtained reduced-dimensional HRSI. Finally, the validity of the UBSF method is verified based on three measured HRSI datasets. The experimental results show that the proposed UBSF is superior to other state-of-the-art BS methods.

Y-Net: Identification of Typical Diseases of Corn Leaves Using a 3D-2D Hybrid CNN Model Combined with a Hyperspectral Image Band Selection Module.

Corn diseases are one of the significant constraints to high-quality corn production, and accurate identification of corn diseases is of great importance for precise disease control. Corn anthracnose and brown spot are typical diseases of corn, and the early symptoms of the two diseases are similar, which can be easily misidentified by the naked eye. In this paper, to address the above problems, a three-dimensional-two-dimensional (3D-2D) hybrid convolutional neural network (CNN) model combining a band selection module is proposed based on hyperspectral image data, which combines band selection, attention mechanism, spatial-spectral feature extraction, and classification into a unified optimization process. The model first inputs hyperspectral images to both the band selection module and the attention mechanism module and then sums the outputs of the two modules as inputs to a 3D-2D hybrid CNN, resulting in a Y-shaped architecture named Y-Net. The results show that the spectral bands selected by the band selection module of Y-Net achieve more reliable classification performance than traditional feature selection methods. Y-Net obtained the best classification accuracy compared to support vector machines, one-dimensional (1D) CNNs, and two-dimensional (2D) CNNs. After the network pruned the trained Y-Net, the model size was reduced to one-third of the original size, and the accuracy rate reached 98.34%. The study results can provide new ideas and references for disease identification of corn and other crops.

Double Deep Q-Network for Hyperspectral Image Band Selection in Land Cover Classification Applications

Hyperspectral data usually consists of hundreds of narrow spectral bands and provides more detailed spectral characteristics compared to commonly used multispectral data in remote sensing applications. However, highly correlated spectral bands in hyperspectral data lead to computational complexity, which limits many applications or traditional methods when applied to hyperspectral data. The dimensionality reduction of hyperspectral data becomes one of the most important pre-processing steps in hyperspectral data analysis. Recently, deep reinforcement learning (DRL) has been introduced to hyperspectral data band selection (BS); however, the current DRL methods for hyperspectral data BS simply remove redundant bands, lack the significance analysis for the selected bands, and the reward mechanisms used in DRL only take basic forms in general. In this paper, a new reward mechanism strategy has been proposed, and Double Deep Q-Network (DDQN) is introduced during BS using DRL to improve the network stabilities and avoid local optimum. To verify the effect of the proposed BS method, land cover classification experiments were designed and carried out to analyze and compare the proposed method with other BS methods. In the land cover classification experiments, the overall accuracy (OA) of the proposed method can reach 98.37%, the average accuracy (AA) is 95.63%, the kappa coefficient (Kappa) is 97.87%. Overall, the proposed method is superior to other BS methods. Experiments have also shown that the proposed method works not only for airborne hyperspectral data (AVIRIS and HYDICE), but also for hyperspectral satellite data, such as PRISMA data. When hyperspectral data is applied to similar applications, the proposed BS method could be a candidate for the BS preprocessing options.

Soft hypergraph regularized weighted low rank subspace clustering for hyperspectral image band selection

ABSTRACT Recently, the graph regularized low-rank representation (GLRR) has been introduced in Hyperspectral Image (HSI) to explore global structures information by exploring the lowest-rank representation of all the data jointly and the local geometrical structure by the graph regularization. However, the traditional graph models are mostly based on a simple intrinsic structure. In this paper, to represent the complex intrinsic band information and further enhance the low rank of the matrix, we propose a soft hypergraph regularized weighted low-rank subspace clustering (HGWLRSC) method for HSI band selection. On the one hand, considering the complex correlation between adjacent bands, hypergraph technique is introduced, which take advantage of the band similarity properties to extract more valuable information and reveals the intrinsic multiple relationships of HSI band sets. On the other hand, the weighted low-rank subspace clustering model is introduced to not only capture the global structure information for the learned representation coefficient matrix but also to consider the importance of different rank components. The proposed algorithm was tested on three widely used hyperspectral data sets, and the experimental results indicate that the proposed HGWLRSC algorithm outperforms the other state-of-the art methods and achieves a very competitive band selection performance for HSI.

Novel hyperbolic clustering-based band hierarchy (HCBH) for effective unsupervised band selection of hyperspectral images

For dimensionality reduction of HSI, many clustering-based unsupervised band selection (UBS) methods have been proposed due to their superiority of reducing the high redundancy between selected bands. However, most of these methods fail to reflect the data structure of HSI, leading to inconsistent results of band selection. To tackle this particular issue, we have proposed a novel hyperbolic clustering-based band hierarchy (HCBH) to fully represent the underlying spectral structure and obtain a more consistent band selection. With the proposed adaptive hyperbolic clustering, the performance can be effectively improved with the aid of geometrical information. By introducing a cluster-centre based ranking metric, the desired band subset can be naturally obtained during the clustering process. Experimental results on three popularly used datasets have validated the superior performance of the proposed approach, which outperforms a few state-of-the-art (SOTA) UBS approaches.

Hyper-graph regularized subspace clustering with skip connections for band selection of hyperspectral image

The Hughes phenomenon of Hyperspectral images (HSIs) with the hundreds of continuous narrow bands makes the computational cost of HSIs processing high. Band selection is an effective way to solve such a problem and a lot of band selection methods have been proposed in recent years. In this paper, a novel hyper-graph regularized subspace clustering with skip connections (HRSC-SC) is proposed for band selection of hyperspectral image, which is a clustering-based band selection method. The networks combine subspace clustering into the convolutional auto-encoder by thinking of it as a self-expressive layer. To make full use of the historical feature maps obtained from the networks and tackle the problem of gradient vanishing caused by multiple nonlinear transformations, the symmetrical skip connections are added to the networks to pass image details from encoder to decoder. Furthermore, the hyper-graph regularization is presented to consider the manifold structure reflecting geometric information within data, which accurately describes the multivariate relationship between data points and makes the results of clustering more accurate so that select the most representative band subset. The proposed HRSC-SC band selection method is compared with the existing robust band selection algorithms on Indian Pines, Salinas-A, and Pavia University HSIs, showing that the results of the proposed method outperform the current state-of-the-art band selection methods. Especially, the overall accuracy of the clustering is the best on three real HSIs compared to other methods when the band selection number is 25, reaching 82.62%, 92.48%, and 96,5% respectively.

Effective band selection of hyperspectral image by an attention mechanism-based convolutional network

The selection of effective and representative spectral bands is extremely important in eliminating redundant information and reducing the computational burden for the potential real-time applications of hyperspectral imaging. However, current band selection methods act as a separate procedure before model training and are implemented merely based on extracted average spectra without incorporating spatial information. In this paper, an end-to-end trainable network framework that combines band selection, feature extraction, and model training was proposed based on a 3D CNN (convolutional neural network, CNN) with the attention mechanism embedded in its first layer. The learned band attention vector was adopted as the basis of a band importance indicator to select effective bands. The proposed network was evaluated by two datasets, a regression dataset for predicting the relative chlorophyll content (soil and plant analyzer development, SPAD) of basil leaves and a classification dataset for detecting the drought stress of pepper leaves. A number of calibration models, including SVM, 1D-CNN, 2B-CNN (two-branch CNN), 3D ResNet and the developed network were established for performance comparison. Results showed that the effective bands selected by the proposed attention-based model achieved higher regression R2 values and classification accuracies not only than the full-spectrum data, but also than the comparative band selection methods, including traditional SPA (successive projections algorithm) and GA (genetic algorithm) methods and the latest 2B-CNN algorithm. In addition, different from the traditional methods, the proposed band selection algorithm can effectively select bands while carrying out model training and can simultaneously take advantage of the original spectral–spatial information. The results confirmed the usefulness of the proposed attention mechanism-based convolutional network for selecting the most effective band combination of hyperspectral images.

Band Selection Using Dilation Distances

In this letter, we adapt the dilation operator from mathematical morphology to propose dilation distances. These dilation distances are then used for band selection in hyperspectral images. It is shown that dilation distances between bands can capture the spatial distance between the objects. Hence, using dilation-based distances would select those bands which identify spatially separated objects. This is illustrated using both toy and real data sets. Furthermore, we compare the proposed approach with existing methods and show empirically that dilation-distance-based band selection provided competitive results outperforming several methods.

A dense spatial–spectral attention network for hyperspectral image band selection

ABSTRACT Hyperspectral images (HSIs) are usually composed of hundreds of bands, which are highly correlated and redundant, leading to dimension disaster and high complexity of classification. In this paper, we propose an end-to-end dense spatial–spectral attention network (DSSAN) for hyperspectral image band selection to reduce the complexity of classification while ensuring the accuracy. In this network, an embeddable spatial–spectral attention module is designed, which can adaptively select the spectral bands from the raw input data. Moreover, this module is a plug-and-play complementary component and embedded in a dense convolutional network (DenseNet) for end-to-end training. The experimental results on two classic hyperspectral data sets demonstrate that the proposed method is superior to several mainstream band selection methods in classification accuracy and the selected band subset has lower redundancy, which can meet the application requirements.

A multi-strategy integrated multi-objective artificial bee colony for unsupervised band selection of hyperspectral images

Abstract As the spectral dimension of hyperspectral images increases, band selection becomes more and more important when using hyperspectral data. Evolutionary algorithms have been applied to the band selection problem of hyperspectral images, but most of existing methods focus only on single indicator, ignoring the whole characteristics of hyperspectral image. In this paper we study a multi-objective artificial bee colony approach for the band selection problem of hyperspectral images. Firstly, a new multi-objective unsupervised band selection model is proposed by using both band correlation and information amount. Secondly, a multi-strategy integrated multi-objective artificial bee colony algorithm (MABC-BS) is proposed to deal with the band selection model above. Several new operators, including the multi-direction search strategy, the x-space crowing degree-based search strategy, and the adaptive mutation, are developed to enhance the proposed algorithm. Compared with eight representative algorithms on three typical test problems, experimental results on the classification performance of four classifiers (i.e., Random Forest, SVM, KNN, ObRaF) show that the proposed algorithm is a powerful approach for tackling the problem of hyperspectral band selection.

A new hybrid SSA-TA: Salp Swarm Algorithm with threshold accepting for band selection in hyperspectral images

Hyperspectral images classification is a primordial step to produce the Land Use maps. Unfortunately, the classification accuracy depends largely on the quality of spectral bands. Several bands are non-informative and the adjacent bands are generally highly correlated. This paper presents a novel band selection approach named SSA-TA based on Salp Swarm Algorithm (SSA) which a new metaheuristic recently developed and Threshold Acceptance(TA). The proposed approach SSA-TA is a hybrid metaheuristic used to select the relevant bands by eliminating the irrelevant and redundant bands to enhance the hyperspectral image classification. This work presents two main ideas. Firstly, we propose a hybridization model based on SSA and Threshold Acceptance (TA). The basic idea is using SSA to find the promising region and use TA to enhance the exploration of the best solution. Secondly, the fitness function is designed to take into consideration three important terms: (1) the maximization of classification accuracy rate (2) the minimization of the number of selected bands (3) the minimization of correlated bands. The performance evaluation of the proposed approach is tested on three hyperspectral images widely used on remote sensing. The proposed approach is compared to other algorithms. The experimental results demonstrate the efficiency of our approach in improving the classification accuracy rate.

Improved quantum evolutionary particle swarm optimization for band selection of hyperspectral image

ABSTRACT The large number of bands, the huge amount of information and the high band correlation in hyperspectral images bring great difficulties to the band selection of hyperspectral images. In the basic Particle Swarm Optimization (PSO), the learning factor and inertia factor are fixed, which limits the exploratory ability of the algorithm and cannot balance the global and local relations well. Therefore, a new algorithm-Improved Quantum Evolutionary Particle Swarm Optimization (IQEPSO) is proposed that the learning factor and inertia factor could be changed with the number of iterations. At the same time, mutation probability could be changed with the number of invariable fitness values, in order to enable particles to jump out of the local optimum. The proposed algorithm is applied to band selection of hyperspectral images. Experiments show that the proposed algorithm can improve the classification accuracy of ground objects. The disadvantage of falling into local optimum is overcome, the convergence speed is accelerated, and better classification accuracy is obtained.

BS-Nets: An End-to-End Framework for Band Selection of Hyperspectral Image

Hyperspectral image (HSI) consists of hundreds of continuous narrow bands with high spectral correlation, which would lead to the so-called Hughes phenomenon and the high computational cost in processing. Band selection has been proven effective in avoiding such problems by removing the redundant bands. However, many of existing band selection methods separately estimate the significance for every single band and cannot fully consider the nonlinear and global interaction between spectral bands. In this paper, by assuming that a complete HSI can be reconstructed from its few informative bands, we propose a general band selection framework, Band Selection Network (termed as BS-Net). The framework consists of a band attention module (BAM), which aims to explicitly model the nonlinear inter-dependencies between spectral bands, and a reconstruction network (RecNet), which is used to restore the original HSI cube from the learned informative bands, resulting in a flexible architecture. The resulting framework is end-to-end trainable, making it easier to train from scratch and to combine with existing networks. We implement two BS-Nets respectively using fully connected networks (BS-Net-FC) and convolutional neural networks (BS-Net-Conv), and compare the results with many existing band selection approaches for three real hyperspectral images, demonstrating that the proposed BS-Nets can accurately select informative band subset with less redundancy and achieve significantly better classification performance with an acceptable time cost.

Hyper-Graph Regularized Kernel Subspace Clustering for Band Selection of Hyperspectral Image

Band selection is an effective way to deal with the problem of the Hughes phenomenon and high computation complexity in hyperspectral image (HSI) processing. Based on the hypothesis that all the pixels are sampled from the union of subspaces, many robust band selection algorithms based on subspace clustering were introduced in recent works, achieving significant performances. However, these methods focus on linear subspaces, which are not suitable for the typical nonlinear structure of HSIs. In this paper, to deal with these obstacles, a new hyper-graph regularized kernel subspace clustering (HRKSC) is presented for band selection of hyperspectral image. The proposed approach extends subspace clustering to nonlinear manifold by utilizing the kernel trick, which can better fit the nonlinear structure of HSIs. The hyper-graph regularized is introduced to consider the manifold structure reflecting geometric information and accurately describe the multivariate relationship between data points, which makes the modeling of HSIs more accurate. The results of the proposed algorithm are compared with existing band selection methods on three well-known hyperspectral data sets, showing that the HRKSC algorithm can accurately select an informative band subset and outperforming the current state-of-the-art band selection methods.

Efficient Graph Convolutional Self-Representation for Band Selection of Hyperspectral Image

Hyperspectral image (HSI) band selection (BS) is an important task for HSI dimensionality reduction, whose goal is to select an informative band subset containing less redundancy. However, traditional BS methods basically work in the Euclidean domain, and thus, often neglect to consider the structural information of spectral bands. In this article, to make full use of the structural information, a novel BS method termed as efficient graph convolutional self-representation (EGCSR) is proposed by incorporating graph convolution into the self-representation model. Since the proposed method is typically modeled in the non-Euclidean domain, it tends to result in a more robust self-representation coefficient matrix. We provide a closed-form solution to the EGCSR model, which leads to high-computational efficiency. We further propose two strategies to determine the informative band subset from the coefficient matrix. The first is a ranking-based strategy, which ranks every band by calculating the cumulative contribution, and the second is a clustering-based strategy, which treats BS as a band clustering task based on using subspace segmentation. Extensive experimental results on three real HSI datasets show that the proposed EGCSR model is dramatically superior to many existing BS methods, and with high-computational efficiency.

Derivative-based band clustering and multi-agent PSO optimization for optimal band selection of hyper-spectral images

Images (HSIs) are popular in diversified applications, such as geosciences, biomedical imaging, molecular biology, agriculture, astronomy, food quality and safety assessment, surveillance and physics-related research. The rich spatial and spectral information of HSI is the key factors for robust representation of class-specific objects, in remote sensing applications. But, these images often suffer from Hughes effect. This effect causes the recording of information about a single scene in multiple spectral bands. This demands a dimensionality reduction step, which can either be a feature reduction/extraction or a feature selection. The feature selection process is commonly called band selection (BS) in the HS data set. The current study proposes an unsupervised BS technique, which is accomplished in three steps, including preprocessing of spectral bands, adjacent band clustering, and multi-agent optimization. Spatio-spectral (using a simple Gaussian filter) information is extracted to evaluate the performance using SVM classifier with different state-of-the-art band selection approaches. The performance of the proposed approach is evaluated for metrics including overall accuracy (OA), average accuracy (AA) and Kappa ( $$\kappa$$ ). The experimental results are promising as these surpass that of other approaches.

Band selection of hyperspectral image by sparse manifold clustering

Band selection of hyperspectral images is an optimal feature selection method, which aims at reducing the computational burden associated with processing the whole data. The significant and informative bands identified by the band selection process lead to efficient, compact representation of the image data and produce a satisfactory performance in the succeeding applications viz. classification, unmixing, target detection and so on. In this study, the authors present an unsupervised manifold clustering approach for band selection, which accounts for different types of scenarios. Unlike other band selection approaches, the authors’ proposed manifold clustering framework identifies the informative bands by utilising the interrelation between the bands and accounts for the multi-manifold structure prevalent in some real images. The proposed band selection framework identifies the optimal number of clusters by cluster validity index, clusters the bands by manifold clustering and select representative bands from each cluster according to graph weight. Their proposed manifold clustering approach is a generic clustering approach, which produces a satisfactory result even when the data contains non-linearity. The information theoretic performance measures, classification and unmixing performance on real image experiments demonstrate the proficiency of their proposed band selection algorithm.

Boltzmann Entropy-Based Unsupervised Band Selection for Hyperspectral Image Classification

Band selection for hyperspectral images helps improve the efficiency of data processing and even the accuracy of classification. It is to reduce the dimensionality of a hyperspectral image by selecting representative bands. In such a process, the quantification of band similarity is the fundamental issue, and it is usually achieved by using an information-theoretic measure, such as mutual information or relative entropy. However, these measures are incapable of quantifying similarity in terms of both composition and configuration. To solve this problem, the Boltzmann entropy (BE), which captures both configurational and compositional information, is employed in this letter. More precisely, the difference in BE between two bands is used for such quantification. The corresponding search strategy is designed for band selection. Experimental evaluation was carried out using remote sensing images for classification. The results clearly demonstrate the superiority of the proposed band selection method over traditional information-theoretic methods: an increase of up to 27% in classification accuracy was observed when using the difference in relative BE with 20 selected bands. In addition, another comparison with some state-of-the-art methods was conducted. The results show that the proposed method is still very competitive; it outperformed all the others when the number of selected bands ranges from 18 to 23. This letter, the first of its kind, reveals that the BE may form a new base for information-theoretic approaches to image processing and even for spatial information science in a broad sense.