Deep learning based hyperspectral band selection using position-sensitive axial attention mechanism

Hyperspectral band selection plays an important role in overcoming the curse of dimensionality. Recently, clustering-based band selection methods have shown promise in the selection of informative and representative bands from hyperspectral images (HSIs). However, most existing clustering-based band selection methods involve the clustering of original HSIs, limiting their performance because of the high dimensionality of hyperspectral bands. To tackle this problem, a novel hyperspectral band selection method termed joint learning of correlation-constrained fuzzy clustering and discriminative non-negative representation for hyperspectral band selection (CFNR) is presented. In CFNR, graph regularized non-negative matrix factorization (GNMF) and constrained fuzzy C-means (FCM) are integrated into a unified model to perform clustering on the learned feature representation of bands rather than on the original high-dimensional data. Specifically, the proposed CFNR aims to learn the discriminative non-negative representation of each band for clustering by introducing GNMF into the model of the constrained FCM and making full use of the intrinsic manifold structure of HSIs. Moreover, based on the band correlation property of HSIs, a correlation constraint, which enforces the similarity of clustering results between neighboring bands, is imposed on the membership matrix of FCM in the CFNR model to obtain clustering results that meet the needs of band selection. The alternating direction multiplier method is adopted to solve the joint optimization model. Compared with existing methods, CFNR can obtain a more informative and representative band subset, thus can improve the reliability of hyperspectral image classifications. Experimental results on five real hyperspectral datasets demonstrate that CFNR can achieve superior performance compared with several state-of-the-art methods.

Hyperspectral remote sensing sensors have the ability to capture a wide range of spectrum of ground objects with hundreds to thousands of bands. The obtained hyperspectral images contain more detailed spectral information than conventional panchromatic or color images. However, redundant and noisy data could also be introduced into the images and may impair further processing of the data. Removing the redundancy and noisy bands becomes one of the most important preprocessing steps of hyperspectral imaging. As a feature selection method, an attention mechanism has been successfully used in computer vision and natural language processing to enable algorithms to concentrate on the most significant data. In this article, we propose a band attention network for hyperspectral image classification. This network can automatically learn to attend to the desired band set that maximizes the classification accuracy. With careful design, the band attention framework can undertake both band weighting and band selection tasks. When working in the band weighting mode, the proposed band attention framework can learn a band weighting vector that models the relationship between all the bands. For band selection, our network can be adapted as two different types of band selection networks that select the significant bands and discard the useless bands. In both tasks, the attention and classification can be learned end to end. The experimental results prove the effectiveness and advantages of the proposed work.

Band selection, as a special case of the feature selection problem, tries to remove redundant bands and select a few important bands to represent the whole image cube. This has attracted much attention, since the selected bands provide discriminative information for further applications and reduce the computational burden. Though hyperspectral band selection has gained rapid development in recent years, it is still a challenging task because of the following requirements: 1) an effective model can capture the underlying relations between different high-dimensional spectral bands; 2) a fast and robust measure function can adapt to general hyperspectral tasks; and 3) an efficient search strategy can find the desired selected bands in reasonable computational time. To satisfy these requirements, a multigraph determinantal point process (MDPP) model is proposed to capture the full structure between different bands and efficiently find the optimal band subset in extensive hyperspectral applications. There are three main contributions: 1) graphical model is naturally transferred to address band selection problem by the proposed MDPP; 2) multiple graphs are designed to capture the intrinsic relationships between hyperspectral bands; and 3) mixture DPP is proposed to model the multiple dependencies in the proposed multiple graphs, and offers an efficient search strategy to select the optimal bands. To verify the superiority of the proposed method, experiments have been conducted on three hyperspectral applications, such as hyperspectral classification, anomaly detection, and target detection. The reliability of the proposed method in generic hyperspectral tasks is experimentally proved on four real-world hyperspectral data sets.

Retrieval of crop biophysical-biochemical variables from airborne AVIRIS-NG data using hybrid inversion of PROSAIL-D

Effective soil spectral band selection and modeling methods can improve modeling accuracy. To establish a hyperspectral prediction model of soil organic matter (SOM) content, this study investigated a forested Eucalyptus plantation in Huangmian Forest Farm, Guangxi, China. The Ranger and Lasso algorithms were used to screen spectral bands. Subsequently, models were established using four algorithms: partial least squares regression, random forest (RF), a support vector machine, and an artificial neural network (ANN). The optimal model was then selected. The results showed that the modeling accuracy was higher when band selection was based on the Ranger algorithm than when it was based on the Lasso algorithm. ANN modeling had the best goodness of fit, and the model established by RF had the most stable modeling results. Based on the above results, a new method is proposed in this study for band selection in the early phase of soil hyperspectral modeling. The Ranger algorithm can be applied to screen the spectral bands, and ANN or RF can then be selected to construct the prediction model based on different datasets, which is applicable to establish the prediction model of SOM content in red soil plantations. This study provides a reference for the remote sensing of soil fertility in forests of different soil types and a theoretical basis for developing portable equipment for the hyperspectral measurement of SOM content in forest habitats.

Hyperspectral image (HSI) classification is one of the growing research areas in Remote Sensing. The high dimensionality of the data cube and a high correlation among the hyperspectral bands, poses a difficulty in HSI processing, and selection is one of the promising solutions to address the issue. In this research paper, a whale optimization-based band selection is proposed. Initially, the informative bands are retrieved via a search mechanism which resembles the hunting behaviour of the humpback whales. Later, a hybrid filter is proposed to extract the intrinsic and edge preserving spatial features. Finally, the smoothened bands are trained using a non-linear support vector machine for efficient classification. The exploitation and exploration in the local and global search for the selection of whale hyperspectral bands from the search space proved to be useful in achieving the high-quality classification maps. The overall accuracy reported on the three benchmark data sets – Indian Pines, University of Pavia and Salinas are 99.05, 97.88, and 97.56%, respectively. The proposed method accuracy is achieved with 5% fewer bands on Indian Pines, 70% on University of Pavia and 50% on Salinas Datasets over the other competing methods, which show the effectiveness of the whale optimization-based band selection.

Hyperspectral band selection (BS) methods relying solely on volume-based criteria often fail to exclude noisy bands and may yield redundant subsets with limited representativeness. To address these limitations, we propose two novel BS algorithms that integrate canonical correlation analysis (CCA) and projection energy measures with determinant-based volume criterion. By modelling the correlation structure between selected and unselected bands, we effectively suppress the inclusion of noisy or outlier bands and enhance both the representativeness and independence of the chosen subsets. Furthermore, we introduce complexity reduction techniques, including the simplified computation of canonical correlation coefficients and recursive updates of inverse matrices, to facilitate efficient implementation for large-scale datasets. Experimental results on five real-world hyperspectral datasets demonstrate that the proposed methods achieve higher classification accuracies, greater robustness against noise, and significantly lower computational costs compared to traditional volume-based and linear algebra-based BS algorithms. These findings underscore the effectiveness and practical value of integrating correlation structure into volume-based BS frameworks.

Multi-agent deep reinforcement learning for hyperspectral band selection with hybrid teacher guide

In this paper, ant colony optimization (ACO) is applied to hyperspectral band selection. The objective is to select a small band subset such that classification accuracy can be maintained or even improved. The ACO-based band selection technique in this research is independent of any classifier, resulting in lower computational cost. Both supervised (i.e., Jeffries-Matusita distance) and unsupervised (i.e., simplex volume) selection criteria are investigated. The experimental results demonstrate that the classification accuracy on the selected bands is higher than using all bands, and ACO-based methods can outperform the widely used sequential forward selection (SFS) method.

Hyperspectral band selection (BS) is an important technique to reduce data dimensionality for the classification applications of hyperspectral remote sensing images (HSIs). Recently, searching-based BS methods have received increasing attention for their ability to select the best subset of bands while preserving the essential information of the original data. However, existing searching-based BS methods neglect effective exploitation of the spatial and spectral prior information inherent in the data, thus limiting their performance. To address this problem, in this study, a novel unsupervised BS method called Spectral–Spatial Iterative Greedy Algorithm (SSIGA) is proposed. Specifically, to facilitate efficient local search using spectral information, SSIGA conducts clustering on all the bands by employing a K-means clustering method with balanced cluster size constraints and constructs a K-nearest neighbor graph for each cluster. Based on the nearest neighbor graphs, SSIGA can effectively explore the neighborhood solutions in local search. In addition, to efficiently evaluate the discriminability and information redundancy of the solution given by SSIGA using the spatial and spectral information of HSIs, we designed an effective objective function for SSIGA. The value of the objective function is derived by calculating the Fisher score for each band in the solution based on the results of the superpixel segmentation performed on the target HSI, as well as by computing the average information entropy and mutual information of the bands in the solution. Experimental results on three publicly available real HSI datasets demonstrate that the SSIG algorithm achieves superior performance compared to several state-of-the-art methods.

MIMR-DGSA: Unsupervised hyperspectral band selection based on information theory and a modified discrete gravitational search algorithm

Selection of useful bands plays a very important role in hyperspectral image classification. In the past decade, metaheuristic algorithms have been used as promising methods for solving this problem. However, many metaheuristic algorithms may provide unsatisfactory performance due to their slow or premature convergence. Therefore, how to develop algorithms well balancing the exploration and exploitation, and find the suitable bands precisely is still a challenge. In this paper, a new hybrid global optimization algorithm, which is based on the Wind Driven Optimization (WDO) and Cuckoo Search (CS) is proposed to solve hyperspectral band selection problems. Both WDO and CS have strong searching ability and require less control parameters, but easily suffer from premature convergence due to loss of diversity of population. The proposed approach uses the Chebyshev chaotic map to initialize the population at initial step. The population is divided into two subgroups and WDO and CS are adopted for these two subgroups independently. By division, these two subgroups can share suitable information and utilize each other’s pros, thus avoid premature convergence, and obtain best optimal solution. Furthermore, the Levy flight step size in CS algorithm is adaptively adjusted based on fitness value and current iteration number, which helps in boosting the convergence speed of algorithm. The experimental results on three standard benchmark datasets namely, Pavia University, Botswana and Indian Pines, prove the superiority of the proposed approach over standard WDO and CS approaches as well as the other traditional approaches in terms of classification accuracy with fewer bands.

When spectral bands are considered as scatter points in the image space, hyperspectral band selection can be transformed into the column subset selection problem for matrices. We consider the problem of finding the submatrix of maximum volume of a matrix <inline-formula xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink"><tex-math notation="LaTeX">$\mathbf {X}\in \mathbb {R}^{N\times L}$</tex-math></inline-formula> <inline-formula xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink"><tex-math notation="LaTeX">$(N\gg L)$</tex-math></inline-formula> within the column space perspective. To solve this problem, we present a novel determinant formula that utilizes the scalar product of arbitrary-order antisymmetric tensors in wedge product form, providing a fresh perspective for interpreting the volume gradient. Based on the variant of this formula, we can mitigate the local extremum problem inherent in existing subset selection algorithms to a certain extent. Subsequently, we introduce exterior algebra theory into the remote sensing community and develop a hyperspectral band selection algorithm via higher order joint volume gradient. To address bias in the volume criterion and reduce algorithmic complexity, we integrate a general subspace partitioning framework that accounts for spectral information. In addition, we propose a method for automatically determining the optimal number of bands. Experiments conducted on both publicly available and Gaofen-5 satellite hyperspectral datasets demonstrate the superiority and stability of our proposed algorithms in terms of classification accuracy and computational efficiency.

A novel dependency definition exploiting boundary samples in rough set theory for hyperspectral band selection

Different crops, as well as the same crop at different growth stages, display distinct spectral and spatial characteristics in hyperspectral images (HSIs) due to variations in their chemical composition and structural features. However, the narrow bandwidth and closely spaced spectral channels of HSIs result in significant data redundancy, posing challenges to crop identification and classification. Therefore, the dimensionality reduction in HSIs is crucial. Band selection as a widely used method for reducing dimensionality has been extensively applied in research on crop identification and mapping. In this paper, a crop superpixel-based affinity propagation (CS-AP) band selection method is proposed for crop identification and mapping in agriculture using HSIs. The approach begins by gathering crop superpixels; then, a spectral band selection criterion is developed by analyzing the variations in the spectral and spatial characteristics of crop superpixels. Finally, crop identification bands are determined through an efficient clustering approach, AP. Two typical agricultural hyperspectral data sets, the Salinas Valley data set and the Indian Pines data set, are selected for validation, each containing 16 crop classes, respectively. The experimental results show that the proposed CS-AP method achieves a mapping accuracy of 92.4% for the Salinas Valley data set and 88.6% for the Indian Pines data set. When compared to using all bands, two unsupervised band selection techniques, and three semi-supervised band selection techniques, the proposed method outperforms others with an improvement of 3.1% and 4.3% for the Salinas Valley and Indian Pines data sets, respectively. Indicate that the proposed CS-AP method achieves superior mapping accuracy by selecting fewer bands with greater crop identification capability compared to the other band selection methods. This research’s significant results demonstrate the potential of this approach in precision agriculture, offering a more cost-effective and timely solution for large-scale crop mapping and monitoring in the future.