Hyperspectral Image Classification Research Articles

Principal space approximation ensemble discriminative marginalized least-squares regression for hyperspectral image classification

Discriminative marginalized least-squares regression (DMLSR) is unable to extract the spectral-spatial joint features, the proportion of learned interfering pixels is high. To solve this problem, a novel principal space approximation ensemble discriminant edge least-squares regression, namely PSAE-DMLSR is proposed for hyperspectral image classification. In the PSAE-DMLSR, a marginal principal component method (MP) is employed to search the optimal spectral subspace, and a principal space local marginal principal component (PSLMP) method is proposed to search the optimal representation space (ORS). In the PSLMP, a principal space representation (PSR) is designed to integrate the global spectral-spatial joint features information of the ORS, and the PSR is used to impose approximate averaging constraints and stochastic cascade fusion on the ORS, which can further improve the representation ability of the ORS. The ORS can effectively reduce the proportion of interfering pixels in DMLSR learning. It conducted comparative experiments with some more advanced classification methods on the three commonly used hyperspectral datasets. The experiment results show that the PSAE-DMLSR classification model can still obtain high classification accuracy under low hardware conditions, and the execution efficiency also has advantages.

Leveraging adaptive spatial shifts multi-layer perceptron with parameterized spatial offsets for hyperspectral image classification

ABSTRACT The architectures based on Multi-Layer Perceptron (MLP) have attracted great attention in hyperspectral image (HSI) classification recently, due to their simplified and efficient architectures. However, such architectures are qualified by the rigid positional relationships between weights and feature elements, inhibiting their capacity to effectively extract diversified features. To address these challenges, An adaptive spatial-shift MLP (AS2MLP) is presented to dynamically modify spatial features by parameterizing learnable spatial offsets. In this way, the AS2MLP can facilitate sample-specific spatial shifts, aligning spatial structures more effectively. Then, An innovative adaptive spatial-shift block (AS2block) is designed to adaptively shift spatial features along distinct spatial axes, enabling the extraction of diversified features separately. It also implements a re-weighting strategy to mitigate redundant features. Building on this foundation, the proposed adaptive spatial-shift network (AS2Net) is for HSI classification. The dual-path AS2Net employs AS2blocks and MLPs for channel mixing, facilitating an adaptive integration of dynamic spatial contextual information dispersed across a range of spectra. The effectiveness of this model is demonstrated using five widely used HSI datasets.

ATSFCNN: a novel attention-based triple-stream fused CNN model for hyperspectral image classification

Abstract Recently, the convolutional neural network (CNN) has gained increasing importance in hyperspectral image (HSI) classification thanks to its superior performance. However, most of the previous research has mainly focused on 2D-CNN, and the limited applications of 3D-CNN have been attributed to its complexity, despite its potential to enhance information extraction between adjacent channels of the image. Moreover, 1D-CNN is typically restricted to the field of signal processing as it ignores the spatial information of HSIs. In this paper, we propose a novel CNN model named attention-based triple-stream fused CNN (ATSFCNN) that fuses the features of 1D-CNN, 2D-CNN, and 3D-CNN to consider all the relevant information of the hyperspectral dataset. Our contributions are twofold: First, we propose a strategy to extract and homogenize features from 1D, 2D, and 3D CNN. Secondly, we propose a way to efficiently fuse these features. This attention-based methodology adeptly integrates features from the triple streams, thereby transcending the former limitations of singular stream utilization. Consequently, it becomes capable of attaining elevated outcomes in the context of hyperspectral classification, marked by increased levels of both accuracy and stability. We compared the results of ATSFCNN with those of other deep learning models, including 1D-CNN, 2D-CNN, 2D-CNN+PCA, 3D-CNN, and 3D-CNN+PCA, and demonstrated its superior performance and robustness. Quantitative assessments, predicated on the metrics of overall accuracy (OA), average accuracy (AA), and kappa coefficient (κ) emphatically corroborate the preeminence of ATSFCNN. Notably, spanning the three remote sensing datasets, ATSFCNN consistently achieves peak levels of OA, quantified at 98.38%, 97.09%, and 96.93% respectively. This prowess is further accentuated by concomitant AA scores of 98.47%, 95.80%, and 95.80%, as well as kappa coefficient values amounting to 97.41%, 96.14%, and 95.21%.

CESA-MCFormer: An Efficient Transformer Network for Hyperspectral Image Classification by Eliminating Redundant Information.

Hyperspectral image (HSI) classification is a highly challenging task, particularly in fields like crop yield prediction and agricultural infrastructure detection. These applications often involve complex image types, such as soil, vegetation, water bodies, and urban structures, encompassing a variety of surface features. In HSI, the strong correlation between adjacent bands leads to redundancy in spectral information, while using image patches as the basic unit of classification causes redundancy in spatial information. To more effectively extract key information from this massive redundancy for classification, we innovatively proposed the CESA-MCFormer model, building upon the transformer architecture with the introduction of the Center Enhanced Spatial Attention (CESA) module and Morphological Convolution (MC). The CESA module combines hard coding and soft coding to provide the model with prior spatial information before the mixing of spatial features, introducing comprehensive spatial information. MC employs a series of learnable pooling operations, not only extracting key details in both spatial and spectral dimensions but also effectively merging this information. By integrating the CESA module and MC, the CESA-MCFormer model employs a "Selection-Extraction" feature processing strategy, enabling it to achieve precise classification with minimal samples, without relying on dimension reduction techniques such as PCA. To thoroughly evaluate our method, we conducted extensive experiments on the IP, UP, and Chikusei datasets, comparing our method with the latest advanced approaches. The experimental results demonstrate that the CESA-MCFormer achieved outstanding performance on all three test datasets, with Kappa coefficients of 96.38%, 98.24%, and 99.53%, respectively.

Artificial gorilla troops optimizer enfolded broad learning system for spatial-spectral hyperspectral image classification

The key challenge of hyperspectral image (HSI) classification is to develop effective feature representations for reducing intra-class variations while expanding inter-class differences. It is of great interest to exploit texture information in the preprocess stage for HSI classification by using feature extraction operators. Additionally, deep learning (DL) networks have focused on spatial-spectral information extraction. Nevertheless, DL-based approaches are time-consuming to compute and difficult to understand owing to the large number of network parameters and insufficient training data. Responding to these issues, this paper introduces a novel spatial-spectral classification framework based on local optimal oriented pattern (LOOP), artificial gorilla troops optimizer (GTO) enfolded broad learning system (BLS), and bi-exponential edge-preserving smoother (LGBB), which mainly includes the following parts: First, the principal component analysis is used to shrink HSI’s spectral dimensions, and then LOOP extracts the rotation-invariant texture-based spatial signatures from the principal components; Next, GTO, a novel nature-inspired and gradient-free optimization algorithm, is first introduced to BLS for finding the global optimum feature subsets, and then LOOP features are fed into the improved BLS to produce soft-classified probability maps; Finally, post-processing filtering is conducted to optimize the soft-classified probability maps by considering spatial contextual information. Experiment results on four actual HSIs demonstrate that the LGBB framework outperforms previous state-of-the-art approaches in quantitative and qualitative outcomes, especially when training samples are constrained.

A hyperspectral image classification method based on feature enhancement and a hybrid deformable convolution network

ABSTRACT In recent years, some hyperspectral image (HSI) classification methods based on deep models have shown excellent performance. Most deep models receive three-dimensional (3D) block structures as input to extract spectral-spatial features from HSI data. However, the existence of background pixel information in 3D blocks affects the spectral-spatial feature representation ability. In addition, directly performing overall dimension reduction on high-dimensional spectral information lacks analysis of the original data, which weakens the constraint relationship between spectral data and increases the difficulty of feature extraction. Therefore, an HSI classification method based on feature enhancement and hybrid deformable convolution network is proposed. Firstly, a novel 3D block is constructed to enhance spectral-spatial feature representation ability by replacing background pixel information with non-background pixel information. Secondly, based on the analysis results of the original spectral data, this paper proposes a regional dimension reduction mechanism to enhance the constraint relationship between HSI data. Finally, the proposed method is applied to hybrid deformable convolution network for experimental verification. The experimental results of three standard datasets show that the classification effect of the proposed method is better than that of several classification methods, which proves the effectiveness of the proposed method.

Hyperspectral image classification using Walsh Hadamard transform- based key band selection and deep convolutional neural networks

ABSTRACT Hyperspectral images (HSIs) have transformed the field of remote sensing by providing researchers with a wealth of information about the Earth’s surface. However, analyzing these images can be an overwhelming task due to the presence of overlapping areas, nested regions, and large intra-class variability. Hyperspectral image classification (HSIC) is a crucial part of identifying the various land cover classes present in hyperspectral images. In order to enhance the accuracy of HSIC, researchers utilize the potential of three-dimensional convolutional neural networks (3D-CNNs). With the ability to influence both the spectral and spatial data present in HSIs, 3D-CNNs provide a promising solution to overcome the challenges associated with HSIC. In this paper, a new method for key band selection is proposed to improve the performance of 3D-CNN model. The proposed method selects the most relevant key bands based on Walsh-Hadamard kernel strength features. These key bands are then used to extract overlapping 3D spatial patches, which serve as input to the proposed 3D-CNN model. To evaluate the performance of the 3D-CNN model six standard benchmark datasets are used. The effectiveness of the proposed method improves the performance of 3D-CNN for HSIC.

Hyperspectral image classification using a double-branch hierarchical partial convolution network

ABSTRACT Recently, convolutional neural network (CNN) and Transformer have been successfully applied in hyperspectral image (HSI) classification. Despite their success, some challenges persist. Popular HSI classification methods typically employ only a single patch as the network input, which may not provide robust information for various datasets or within different scenarios of a specific dataset. Large-scale patches are usually conducive to classifying large-scale scenes, while small-scale patches excel at classifying boundary locations. In addition, the majority of methods do not distinguish between the central pixel and its adjacent pixels, although the information from the central position is crucial. In this paper, a novel double-branch hierarchical partial convolution network (DBHP-Net) is proposed to address the aforementioned challenges. Concretely, DBHP-Net employs a new convolution, named TConv, to emphasize mining the spatial-channel features of the central position. Then, the TConvFormer Block is built by adopting the advanced engineering techniques of Transformer. Furthermore, we construct our model in a hierarchical style. The double-branch structure with large-scale and small-scale patches provides complementary information from the dataset. Experimental results based on randomly selected samples and spatially disjoint samples of two benchmark datasets demonstrate that our method outperforms other state-of-the-art (SOTA) methods.

Multi-Level Feature Extraction Networks for Hyperspectral Image Classification

Hyperspectral image (HSI) classification plays a key role in the field of earth observation missions. Recently, transformer-based approaches have been widely used for HSI classification due to their ability to model long-range sequences. However, these methods face two main challenges. First, they treat HSI as linear vectors, disregarding their 3D attributes and spatial structure. Second, the repeated concatenation of encoders leads to information loss and gradient vanishing. To overcome these challenges, we propose a new solution called the multi-level feature extraction network (MLFEN). MLFEN consists of two sub-networks: the hybrid convolutional attention module (HCAM) and the enhanced dense vision transformer (EDVT). HCAM incorporates a band shift strategy to eliminate the edge effect of convolution and utilizes hybrid convolutional blocks to capture the 3D properties and spatial structure of HSI. Additionally, an attention module is introduced to identify strongly discriminative features. EDVT reconfigures the organization of original encoders by incorporating dense connections and adaptive feature fusion components, enabling faster propagation of information and mitigating the problem of gradient vanishing. Furthermore, we propose a novel sparse loss function to better fit the data distribution. Extensive experiments conducted on three public datasets demonstrate the significant advancements achieved by MLFEN.

Spectral-Spatial Domain Attention Network for Hyperspectral Image Few-Shot Classification

Recently, many deep learning-based methods have been successfully applied to hyperspectral image (HSI) classification. Nevertheless, training a satisfactory network usually needs enough labeled samples. This is unfeasible in practical applications since the labeling of samples is time-consuming and expensive. The target domain samples that need to be classified are usually limited in HSIs. To mitigate this issue, a novel spectral-spatial domain attention network (SSDA) is proposed for HSI few-shot classification, which can transfer the learned classification knowledge from source domain contained enough labeled samples to target domain. The SSDA includes a spectral-spatial module, a domain attention module, and a multiple loss module. The spectral-spatial module can learn discriminative and domain invariance spectral-spatial features. The domain attention module can further enhance useful spectral-spatial features and avoid the interference of useless features. The multiple loss module, including few-shot loss, coral loss, and mmd loss, can solve the domain adaptation issue. Extensive experimental results demonstrate that on the Salinas, the University of Pavia (UP), the Indian Pines (IP), and the Huoshaoyun datasets, the proposed SSDA obtains higher classification accuracies than state-of-the art methods in the HSI few-shot classification.

Dual-input ultralight multi-head self-attention learning network for hyperspectral image classification

ABSTRACT In the hyperspectral image (HSI) classification tasks, various deep learning models have achieved remarkable success. However, most deep learning models are compute-intensive, requiring significant computing power, time, and other resources. It becomes a challenge to pursue better results while saving computational resources. Therefore, a novel dual-input ultralight multi-head self-attention learning network (DUMS-LN) is proposed for HSI classification. The proposed DUMS-LN consists of three main core modules, namely the high-dimensional reduced module (HDRM), lightweight multi-head self-attention (LMHSA) module, and linearized hierarchical conversion module (LHCM). HDRM is used as a pre-processing module with efficient data compression and combines spatial and spectral information extraction from the raw data to provide cleaner and more comprehensive feature data for subsequent processing. In addition, the core computational module of DUMS-LN is the LMHSA module, which is lightweight but possesses better data processing capability than the traditional multi-head self-attention module. Finally, the LHCM divides the model into two phases, reducing the dimensionality of the feature data phase by phase so that the LMHSA module can perform feature extraction at different levels. Experiments on four benchmark HSI datasets show that the proposed DUMS-LN outperforms the comparison HSI classification algorithms regarding speed and classification accuracy.

Relaxation linear regression with spectral–spatial constrained locality adaptive regularization for hyperspectral image classification

AbstractRecently, relaxation linear regression has received increasing attention in image analysis. However, the current relaxation linear regression methods fail to consider the local geometrical structure and spatial information while they are applied for hyperspectral image (HSI) classification. To address the above problems, this letter proposes a novel relaxation linear regression with spectral–spatial constrained locality adaptive regularization (SSLA‐RLR) method for HSI classification. The SSLA‐RLR method not only integrates the locality adaptive graph with relaxation linear regression to adaptively exploit the local geometrical structure for relieving the side effects of noise corruptions, but also takes the spatial correlations between each data point and its neighbours into consideration via spectral–spatial constraint. Finally, extensive experiments are conducted on three benchmark HSI datasets to demonstrate the effectiveness of the proposed method.

Synergistic spectral and spatial feature analysis with transformer and convolution networks for hyperspectral image classification

Synergistic spectral and spatial feature analysis with transformer and convolution networks for hyperspectral image classification

Graph Information Aggregation Cross-Domain Few-Shot Learning for Hyperspectral Image Classification.

Most domain adaptation (DA) methods in cross-scene hyperspectral image classification focus on cases where source data (SD) and target data (TD) with the same classes are obtained by the same sensor. However, the classification performance is significantly reduced when there are new classes in TD. In addition, domain alignment, as one of the main approaches in DA, is carried out based on local spatial information, rarely taking into account nonlocal spatial information (nonlocal relationships) with strong correspondence. A graph information aggregation cross-domain few-shot learning (Gia-CFSL) framework is proposed, intending to make up for the above-mentioned shortcomings by combining FSL with domain alignment based on graph information aggregation. SD with all label samples and TD with a few label samples are implemented for FSL episodic training. Meanwhile, intradomain distribution extraction block (IDE-block) and cross-domain similarity aware block (CSA-block) are designed. The IDE-block is used to characterize and aggregate the intradomain nonlocal relationships and the interdomain feature and distribution similarities are captured in the CSA-block. Furthermore, feature-level and distribution-level cross-domain graph alignments are used to mitigate the impact of domain shift on FSL. Experimental results on three public HSI datasets demonstrate the superiority of the proposed method. The codes will be available from the website: https://github.com/YuxiangZhang-BIT/IEEE_TNNLS_Gia-CFSL.

Near-infrared hyperspectral circular polarization imaging and object classification with machine learning.

We constructed a hyperspectral circular polarization (S3) imaging system in the near-infrared (NIR) region comprising a circularly polarized broadband light source, a polarization grating, and a commercial hyperspectral camera. With this system, we captured hyperspectral S3 images of plastic samples. We then demonstrated the classification with machine learning and found that the hyperspectral S3 images showed higher classification precision than the conventional NIR hyperspectral images. This result indicates that the hyperspectral S3 imaging has potential for object classification even for samples with similar absorption spectra. This hyperspectral S3 imaging system can be applied in garbage classification in recycling plants.

Minimum Noise Fraction and Long Short-Term Memory Model for Hyperspectral Imaging

In recent years, deep learning techniques have presented a major role in hyperspectral image (HSI) classification. Most commonly Convolutional Neural Network (CNN) and Long Short-Term Memory (LSTM) has greatly advanced the accuracy of hyperspectral image classification, making it powerful tool for remote sensing applications. Deep structure learning, which involves multiple layers of neural network, has shown promising results in effectively addressing nonlinear problems and improving classification accuracy and reduce execution time. The exact categorization of ground topographies from hyperspectral data is a crucial and current research topic that has gotten a lot of attention. This research work focuses on hyperspectral image categorization utilizing several machine learning approaches such as support vector machine (SVM), K-Nearest Neighbour (KNN), CNN and LSTM. To reduce the number of superfluous and noisy bands in the dataset, Principal Component Analysis (PCA) and Minimum Noise Fraction (MNF) were utilized. Different performance evaluation measures like time taken for testing, classification accuracy, kappa accuracy, precision, recall, specificity, F1_score, and Gmean have been taken to prove the efficacy of the models. Based on the simulation results, it is observed that the LSTM model outperforms the other models in terms of accuracy percentage and time consumption, making it the most effective model for classifying hyperspectral imaging datasets.

Joint Classification of Hyperspectral Images and LiDAR Data Based on Dual-Branch Transformer

In the face of complex scenarios, the information insufficiency of classification tasks dominated by a single modality has led to a bottleneck in classification performance. The joint application of multimodal remote sensing data for surface observation tasks has garnered widespread attention. However, issues such as sample differences between modalities and the lack of correlation in physical features have limited the performance of classification tasks. Establishing effective interaction between multimodal data has become another significant challenge. To fully integrate heterogeneous information from multiple modalities and enhance classification performance, this paper proposes a dual-branch cross-Transformer feature fusion network aimed at joint land cover classification of hyperspectral imagery (HSI) and Light Detection and Ranging (LiDAR) data. The core idea is to leverage the potential of convolutional operators to represent spatial features, combined with the advantages of the Transformer architecture in learning remote dependencies. The framework employs an improved self-attention mechanism to aggregate features within each modality, highlighting the spectral information of HSI and the spatial (elevation) information of LiDAR. The feature fusion module based on cross-attention integrates deep features from two modalities, achieving complementary information through cross-modal attention. The classification task is performed using jointly obtained spectral and spatial features. Experiments were conducted on three multi-source remote sensing classification datasets, demonstrating the effectiveness of the proposed model compared to existing methods.

Multi-scale feature learning via residual dynamic graph convolutional network for hyperspectral image classification

ABSTRACT Thanks to the relevant spectral-spatial information, hyperspectral images (HSIs) have been widely exploited in Earth observation. Recently, graph convolutional networks (GCNs) have attracted increasing attention in HSI classification due to their advantages in processing non-Euclidean structure data. Unlike convolutional neural networks (CNNs), which perform convolution operations on regular square regions, GCNs can directly work on graph structure data to extract the relationships among adjacent land covers. However, extracting meaningful and deep discriminative spectral-spatial features from HSIs is still a challenging task. In this article, a novel multi-scale feature learning via residual dynamic graph convolutional network is designed for HSI classification, which can extract large-scale contextual spatial structures at superpixel-level graph and local spectral-spatial information at pixel-level, significantly improving the performance of HSI classification. Unlike from the existing GCN-based methods that operate on a graph with a fixed neighbourhood size, multiple graphs with diverse neighbourhood scales are built to comprehensively leverage spectral-spatial information and relationship at multiple scales, and these graphs are dynamically updated to generate more discriminative features (via dynamic GCN) during the convolution process. Moreover, to fully use the multi-scale features extracted from HSIs, a multi-scale feature fusion module is developed to emphasize important features and suppress irrelevant ones. Extensive experiments carried on three benchmark data sets demonstrate the superiority of the proposed approach over other state-of-the-art methods.

Effective hyperspectral image classification based on segmented PCA and 3D-2D CNN leveraging multibranch feature fusion

ABSTRACT We present an innovative hyperspectral image (HSI) classification method addressing challenges posed by closely spaced wavelength bands. Our approach combines 3D-2D convolutional neural networks (CNNs) with multi-branch feature fusion for improved spectral-spatial feature extraction. Using segmented principal component analysis (Seg-PCA), we reduce HSIs’ spectral dimensions into global and local intrinsic characteristics. The integration of 3D and 2D CNNs captures joint spectral-spatial features, while a multi-branch network extracts and merges diverse local features along the spectral dimension. Our method outperforms existing approaches, achieving remarkable accuracy of 99.27%, 100%, and 99.99% on Indian Pines, Salinas Scene, and University of Pavia datasets, respectively.

A novel model for hyper spectral image enhancement and classification: PCA, MBAO and CNN integration

A novel model for hyper spectral image enhancement and classification: PCA, MBAO and CNN integration