Band Selection Method Research Articles

Ensefgram: An optimal demodulation band selection method for the early fault diagnosis of high-speed train bearings

Early fault diagnosis of high-speed train bearings under road spectrum shocks is challenging. As a conventional narrowband demodulation method, the Fast Kurtogram (FK) class of algorithms have some drawbacks: the fixed-band segmentation strategy makes it difficult for the method to find the optimal demodulation frequency band (ODFB); other major limitations are the poor robustness of the statistical index of the blind algorithm and the poor adaptiveness of the target algorithm given the need to the know the bearing operating state in advance. To address these issues, this paper proposes Ensefgram, an early fault detection method for high-speed train bearings with the acoustic emission signal as an example. First, a filter bank is designed by combining the spectral trend with the median dichotomy strategy, which adaptively segments the frequency band and retains the integrity of the sub-band fault components to the maximum extent. In addition, a novel statistical index, termed the maximum weighted spectral energy frequency factor, is proposed, and it is combined with the envelope spectrum negentropy (ES-negentropy) to form a composite index ENSEF, which can select the ODFB blindly. Finally, the proposed method was validated by simulation and applying test signals close to the real working conditions of a high-speed train. Compared with FK, Infogram, and other methods, Ensefgram provides more accurate detection results for the early fault diagnosis of bearings.

Anomaly‐background separation and particle swarm optimization based band selection for hyperspectral anomaly detection

AbstractAs one of the dimensionality reduction techniques of hyperspectral image (HSI), band selection (BS) does not change the spectral characteristics and physical meaning of HSIs, which is beneficial to the identification and analysis of surface objects. Recently, many BS methods for target detection have achieved promising results by making full use of the priori spectral features of the target to be detected. Conversely, anomaly detection separates the anomaly based solely on the statistical distribution difference between anomaly and background without any prior information. Therefore, the development of BS for anomaly detection has lagged far behind that of BS for target detection. To this end, this paper proposes a novel BS algorithm dedicated to anomaly detection tasks, named anomaly‐background separation and particle swarm optimization (PSO)‐based BS. Specifically, an anomaly‐background separation framework (ABSF) is established to predetermine a priori knowledge of anomaly distribution. Then, three band prioritization criteria are constructed with the anomaly‐background constraints generated by ABSF. Finally, PSO is used to find the optimal subset of bands in the solution space. The experiments on two real datasets demonstrate that the proposed method yields better detection results and greater stability compared to other BS methods discussed in this paper.

A novel framework combining band selection algorithm and improved 3D prototypical network for tree species classification using airborne hyperspectral images

Fine-grained identification of forest types and tree species represents a critical aspect of forest resource inventory and monitoring. The use of airborne hyperspectral remote sensing imagery stands out for its ability to finely differentiate among tree species, leveraging its exceptional spatial resolution and rich spectral details. However, this approach is limited by several challenges (e.g., high spectral correlation and information redundancy). In accordance, the adoption of a lightweight deep learning approach in the form of a few-shot learning model can effectively resolve the challenges of multi-forest tree species classification. Therefore, integrating a data dimensionality reduction algorithm with a few-shot classification model presents a promising avenue for resolving the fine-grained classification of forest tree species. In this study, we propose the innovative classification framework FAST 3D-CNN P-Net. This framework utilizes CNN for band selection, enhances the fine-grained identification process in hyperspectral data, and integrates an optimized FAST 3D-CNN into the P-Net classifier (a few-shot classifier). First, a CNN-based band selection method is employed to learn the nonlinear dependencies between spectral bands, assign weights to rank the bands, and reconstruct the global spectral information using the most informative bands. It then constructs a novel classification model, designated FAST 3D-CNN P-Net, through the integration of an optimal 3D-CNN with a prototypical network. To enhance classification performance, the FAST 3D-CNN P-Net utilizes reconstructed hyperspectral images derived from the band selection results as input. The effectiveness of the proposed framework was assessed with the airborne GFF dataset and the widely accessible medium-resolution hyperspectral datasets, Indian Pines (IP) and Kennedy Space Center (KSC). The overall classification accuracy reached 98.33 % for the GFF dataset and 97.21 % and 99.43 % for the IP and KSC, respectively, exhibiting performance superiority compared to the standalone 3D-CNN classification network. This classification framework demonstrates efficiency in selecting a subset of hyperspectral bands with minimal redundancy, empowering the rapid and accurate classification and mapping of tree species in complicated, multi-species forest stands, even with a limited quantity of labeled samples.

Band Selection via Band Density Prominence Clustering for Hyperspectral Image Classification

Band clustering has been widely used for hyperspectral band selection (BS). However, selecting an appropriate band to represent a band cluster is a key issue. Density peak clustering (DPC) provides an effective means for this purpose, referred to as DPC-based BS (DPC-BS). It uses two indicators, cluster density and cluster distance, to rank all bands for BS. This paper reinterprets cluster density and cluster distance as band local density (BLD) and band distance (BD) and also introduces a new concept called band prominence value (BPV) as a third indicator. Combining BLD and BD with BPV derives new band prioritization criteria for BS, which can extend the currently used DPC-BS to a new DPC-BS method referred to as band density prominence clustering (BDPC). By taking advantage of the three key indicators of BDPC, i.e., cut-off band distance bc, k nearest neighboring-band local density, and BPV, two versions of BDPC can be derived called bc-BDPC and k-BDPC, both of which are quite different from existing DPC-based BS methods in three aspects. One is that the parameter bc of bc-BDPC and the parameter k of k-BDPC can be automatically determined by the number of clusters and virtual dimensionality (VD), respectively. Another is that instead of using Euclidean distance, a spectral discrimination measure is used to calculate BD as well as inter-band correlation. The most important and significant aspect is a novel idea that combines BPV with BLD and BD to derive new band prioritization criteria for BS. Extensive experiments demonstrate that BDPC generally performs better than DPC-BS as well as many current state-of-the art BS methods.

The LFIgram: A Targeted Method of Optimal Demodulation Band Selection for Compound Faults Diagnosis of Rolling Bearing

The LFIgram: A Targeted Method of Optimal Demodulation Band Selection for Compound Faults Diagnosis of Rolling Bearing

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.

Multigraph approximate-representation learning for hyperspectral band selection

ABSTRACT Unsupervised hyperspectral image (HSI) band selection methods have been attracting ever-increasing attention. However, the local structural features captured by most of the existing methods suffer from a certain degree of risk of feature ambiguity. Moreover, these methods typically have difficulty preserving the latent manifold structure of the HSI in selected subsets of the bands. To address these problems, firstly, the graph approximate representation learning (GARL) model is proposed in this article by incorporating graph-regularized sparse coding into self-representation learning to preserve the local manifold structure information of HSI in the selected subset of bands. Meanwhile, to alleviate the ambiguity of the local graph structures, we design a local-global feature relationship expression model to construct the latent long-distance contextual connectivity (LDCC) graphs between the obtained local graph structures. Then, to preserve HSI’s local and non-local manifold structure information as much as possible in the selected subset of bands, a multigraph approximate representation learning (MGARL) model is proposed in this article by incorporating the obtained LDCC graph into the proposed GARL model. Next, we design a solution method to solve the proposed formulation. Finally, extensive experimental results on four datasets reveal the promising performance of MGARL over some state-of-the-art methods.

Nondestructive discrimination of Potentilla anserina L. from different production areas based on near-infrared spectroscopy.

The traditional Chinese medicine (TCM) Potentilla anserina L. can use both as food and medicine. At present, the market mainly depends on experience to identify the species and determine the production areas of P. anserina. To ensure the quality of P. anserina, it is essential to improve the level of quality control. We aimed to establish a rapid and nondestructive discrimination model to identify P. anserina from different production areas by near-infrared spectroscopy. The spectra of complete P. anserina medicinal materials and their powder of the same variety from four production areas were collected, and principal component analysis discriminant analysis and partial least squares discriminant analysis (PLS-DA) were conducted based on different pretreatment methods and band selection methods. Then, the spectra of complete medicinal materials were converted into the spectra of medicinal powder for nondestructive identification. The correct recognition rate (CRR) of the PLS-DA discriminant model was the best after spectral preprocessing using autoscaling and competitive adaptive reweighted sampling for band selection. The CRRs of the calibration set and validation set were 100%, the CRRs of the external test set were 95%, 90%, 82%, and 88%, respectively, and the CRRs of the transfer external test set were 84%, 80%, 82%, and 86%, respectively. We realized the nondestructive and effective identification of P. anserina from different origins and laid a foundation for the industrialization and upgrading of TCM.

DTMSgram: a novel optimal demodulation frequency band selection method for wheelset bearings fault diagnosis under wheel-rail excitation

Due to the influence of the wheel-rail excitation and complex transmission path, the fault signature of wheelset bearings is often obscured by complex background noise, which brings great challenges to the adaptive determination of the informative frequency band (IFB) in envelope analysis. In this paper, the vibration response characteristics of the axle box under wheel–rail excitation are revealed through full-scale bench tests. The experimental results show that tread damage will provoke periodic transient impacts and it has an obvious sparsity in the frequency domain. Inspired by this feature, a DTMSgram method is proposed to enhance the fault feature components of vibration signals through time-domain and frequency-domain noise reduction technology, and improve the accuracy of demodulation frequency band selection. Firstly, the amplitude spectrum of different weight coefficients is used to preprocess the vibration signal and adjust the full band component of the vibration signal. Then, the autocorrelation process is performed on each layer’s narrowband filtered signal envelope in kurtogram to further reduce noise interference from a time-domain perspective. Moreover, a two-dimensional color map is constructed showing the normalized squared envelope spectrum of the IFBs determined simultaneously from the modified signal. Finally, the effectiveness of the proposed method is verified by simulated signal and experimental data of three sets of full-size wheelset bearing systems. The analysis results indicate this proposed method can effectively overcome the influence of complex wheel–rail excitation interference and can diagnose multi-source faults simultaneously.

Refined fire detection and band selection method in hyperspectral remote sensing imagery based on sparse-VIT

Hyperspectral remote sensing has powerful ability to perceive information of land surface, and can be applied to the field of fire monitoring to achieve refined fire analysis. In this paper, we study the fire detection method based on hyperspectral remote sensing, design an end to end fire detection model based on sparse visual transformer, and propose a sensitive band selection method under the transformer framework. In order to suppress the interference of invalid bands in hyperspectral data, the sparse attention mechanism and the top-k selection mechanism are introduced in the model to zeroing the attention score of invalid bands. To achieve dimension reduction, a non-maximum attention suppression algorithm under the transformer framework is designed and band pruning is further introduced to eliminate invalid bands and redundant bands. The model adopts input mode of band-exclusive-token, which makes the pruning operation equivalent to band selection. We publish a hyperspectral fire detection dataset and the performance of our proposed model is verified on the dataset.

Self-Supervised Deep Multi-Level Representation Learning Fusion-Based Maximum Entropy Subspace Clustering for Hyperspectral Band Selection

As one of the most important techniques for hyperspectral image dimensionality reduction, band selection has received considerable attention, whereas self-representation subspace clustering-based band selection algorithms have received quite a lot of attention with good effect. However, many of them lack the self-supervision of representations and ignore the multi-level spectral–spatial information of HSI and the connectivity of subspaces. To this end, this paper proposes a novel self-supervised multi-level representation learning fusion-based maximum entropy subspace clustering (MLRLFMESC) method for hyperspectral band selection. Firstly, to learn multi-level spectral–spatial information, self-representation subspace clustering is embedded between the encoder layers of the deep-stacked convolutional autoencoder and its corresponding decoder layers, respectively, as multiple fully connected layers to achieve multi-level representation learning (MLRL). A new auxiliary task is constructed for multi-level representation learning and multi-level self-supervised training to improve its capability of representation. Then, a fusion model is designed to fuse the multi-level spectral–spatial information to obtain a more distinctive coefficient matrix for self-expression, where the maximum entropy regularization (MER) method is employed to promote connectivity and the uniform dense distribution of band elements in each subspace. Finally, subspace clustering is conducted to obtain the final band subset. Experiments have been conducted on three hyperspectral datasets, and the corresponding results show that the proposed MLRLFMESC algorithm significantly outperforms several other band selection methods in classification performance.

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

Tracking moisture contents in the pollution layer on a composite insulator surface using hyperspectral imaging technology.

Electrical insulators used in transmission lines and outdoor substations are exposed to severe environmental pollution, which significantly increases the risk of power system failure, especially when the pollution layer is highly humid due to adverse weather conditions. The focus of this paper is to establish an effective method for assessing the moisture content (MC) in pollution layers as it serves as a crucial indicator for evaluating the risk of failure in insulators. Hyperspectral imaging (HSI) technology with a spectral range of 371.08-1037.89 nm was applied to determine significant changes in reflectance spectral characteristics in insulators during dynamic wetting and drying periods. Partial least squares regression (PLSR) models were utilized to evaluate the data presentation enhancement abilities of spectral transformation models and the data dimensionality reduction abilities of characteristic band selection methods. Furthermore, PLSR models were developed to calculate the MC along the pixel dimension to visually retrieve the dynamic wetting and drying processes of the pollution layer. The R-squared and root-mean-square error (RMSE) results in the cross-verification set and prediction set of the RE-RF(70%)-PLSR model with two characteristic bands with a wavelength of 543.28 nm and 848.01 nm were as follows: RCV2 = 0.9824, RMSECV = 0.0367, RP2 = 0.9818, RMSEP = 0.0369, respectively. This research contributes towards the visualization retrieval of the MC and offers an important technique for analyzing flashover evolution, optimizing insulator design, and preparing coating materials for insulators.

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.

An Unsupervised Band Selection Method via Contrastive Learning for Hyperspectral Images

Band selection (BS) is an efficacious approach to reduce hyperspectral information redundancy while preserving the physical meaning of hyperspectral images (HSIs). Recently, deep learning-based BS methods have received widespread interest due to their ability to model the nonlinear relationship between bands, with existing methods typically relying on generative algorithms. However, the process of generating images with pixel-level detail required by generative algorithm-based BS methods is computationally expensive. To alleviate this issue, we propose a contrastive learning-based unsupervised BS architecture, termed ContrastBS, in this article. With the help of contrastive learning, the proposed architecture avoids the costly generation step in pixel space by learning to distinguish data at the abstract semantic level of the feature space. Specifically, ContrastBS combines an attention mechanism with contrastive learning to extract the importance of each band. Furthermore, we design a novel loss function, which is able to constrain the symmetric loss while ensuring attention to the most valuable bands, for the contrastive learning-based BS network. Experimental results indicate that ContrastBS has excellent classification performance and competitive time cost compared to the comparison methods.

Potentiality of ultraspectral sensor in biophysical and biochemical vegetation parameter inversion

ABSTRACT Biophysical and biochemical vegetation parameters are important indicators for quantitatively monitoring vegetation growth status and ecosystem. Compared to hyperspectral data, the emergence of ultraspectral remote sensing technology supports the acquisition of thousands of more narrow bands in near-infrared spectrum. It is helpful to capture more detailed vegetation spectral features, thereby providing more accurate quantitative inversion. However, there are still many unknowns in utilizing ultraspectral sensors for vegetation biochemical and biophysical parameters inversion, particularly in extracting meaningful inversion information from thousands of spectral bands. Here, we discuss the potentiality, efficiency and accuracy of the ultraspectral sensor AisaIBIS in this field. Firstly, we conducted sensitivity analysis based on simulated sensor dataset and the PROSAIL model to determine the vegetation parameters suitable for inversion. Secondly, in order to enhance the inversion efficiency and reduce the effects of multicollinearity in ultraspectral data, we proposed the stepped asymptotic optimization strategy that combined random forest (RF) and Gaussian process regression (GPR-BAT) methods for dimensionality reduction. Lastly, using experimental data, we compared the inversion accuracy between traditional band selection methods and our proposed method, thereby validating the application potential of ultraspectral sensors. The experimental results indicate that leaf area index (LAI), chlorophyll content, and brown pigment are the most sensitive parameters to the spectral information and are suitable for inversion. Based on the stepped asymptotic optimization strategy, we effectively avoid incomplete information extraction caused by high correlation and spectral redundancy, achieving optimal bands covering the entire spectrum of the sensor. The LAI inversion shows a correlation of 0.691 and an RMSE of 1.955 with the original data, outperforming the accuracy of the traditional band selection methods RF and principal component analysis (PCA). Moreover, the inversion time has decreased by approximately 66% after the dimensionality reduction. This study illustrates the application value of monitoring vegetation parameters by ultraspectral sensor.

An interval band selection method based on class saliency map to identify vegetation under natural gas microleakage stress

When microleakage occurs in underground natural gas pipelines, the surface gas concentration may not reach detectable thresholds. Hyperspectral remote sensing can monitor indirectly through the spectral changes of surface vegetation. However, hyperspectral data often have band redundancy, and existing band selection methods lack consideration for vegetation-specific spectral response characteristics and interpretability. In this study, an interval band selection method based on class saliency map (IBS-CSA) was proposed to extract characteristic bands for identifying vegetation under natural gas microleakage stress. Firstly, a field experiment was designed to investigate the impact of natural gas stress on grass, bean, and wheat. Then, an one-dimensional convolutional neural network was constructed to obtain class saliency maps for different stress levels. Based on these maps, a band importance factor was created, and characteristic bands were extracted by applying the delineated spectral intervals. Finally, the performance of the IBS-CSA method was validated through statistical analysis and comparative modeling. The results indicate that when the proportion of selected bands reaches 10% (equivalent to 10 bands), the characteristic bands selected by the IBS-CSA method perform excellent separability, with JM distances larger than 1.8 between healthy and stressed vegetation. Furthermore, compared to other band selection methods, the IBS-CSA method performs superior accuracy and stability across all classification models, with overall accuracies of 91.40%, 96.67%, and 96.67% on the grass, bean, and wheat test datasets, respectively. This study demonstrates the potential of IBS-CSA in extracting vegetation spectral features under natural gas microleakage stress, providing unique insights into the extraction of characteristic wavelengths of stressed vegetation.

Heterogeneous Regularization-Based Tensor Subspace Clustering for Hyperspectral Band Selection.

Band selection (BS) reduces effectively the spectral dimension of a hyperspectral image (HSI) by selecting relatively few representative bands, which allows efficient processing in subsequent tasks. Existing unsupervised BS methods based on subspace clustering are built on matrix-based models, where each band is reshaped as a vector. They encode the correlation of data only in the spectral mode (dimension) and neglect strong correlations between different modes, i.e., spatial modes and spectral mode. Another issue is that the subspace representation of bands is performed in the raw data space, where the dimension is often excessively high, resulting in a less efficient and less robust performance. To address these issues, in this article, we propose a tensor-based subspace clustering model for hyperspectral BS. Our model is developed on the well-known Tucker decomposition. The three factor matrices and a core tensor in our model encode jointly the multimode correlations of HSI, avoiding effectively to destroy the tensor structure and information loss. In addition, we propose well-motivated heterogeneous regularizations (HRs) on the factor matrices by taking into account the important local and global properties of HSI along three dimensions, which facilitates the learning of the intrinsic cluster structure of bands in the low-dimensional subspaces. Instead of learning the correlations of bands in the original domain, a common way for the matrix-based models, our model learns naturally the band correlations in a low-dimensional latent feature space, which is derived by the projections of two factor matrices associated with spatial dimensions, leading to a computationally efficient model. More importantly, the latent feature space is learned in a unified framework. We also develop an efficient algorithm to solve the resulting model. Experimental results on benchmark datasets demonstrate that our model yields improved performance compared to the state-of-the-art.

A hyperspectral band selection method based on sparse band attention network for maize seed variety identification

The development of a real-time online system for rapid and nondestructive identification of seed varieties can greatly improve production efficiency in modern agriculture. Hyperspectral imaging (HSI) is a powerful tool for seed variety identification. Nevertheless, hyperspectral data are not only high in dimensionality but also contain redundant information, which is very unfriendly to real-time online applications. Selecting a few representative bands from the entire working spectral region can significantly reduce the equipment cost and computational load of HSI. In the field of food and agr-products quality evaluation, Band selection (BS) methods based on chemometrics have been dominant for a long time. Most of these methods, however, fail to take full account of the nonlinearities and global interactions between spectral bands, which may result in the selection of some adjacent bands that still retain more redundant information. In this paper, a novel BS network is proposed, which is composed of sparse band attention module and classification net module. The former is used to generate weight of each band, and sparse constraint is applied to the weights of redundant bands, while the latter is used to achieve high-performance classification with reweighted data. Furthermore, to solve the problem of gradient updating caused by sparse constraint, a auxiliary loss function is defined to assist optimization. Finally, comparative experiments is conducted on our maize seed hyperspectral dataset. The results demonstrate that the presented method selects a subset of informative bands with less redundant information to obtain better classification performance and outperforms several other existing BS methods.

Explainable AI-driven wavelength selection for hyperspectral imaging of honey products

Hyperspectral imaging (HSI) offers a rapid and non-destructive method for inspecting various food and agricultural products. Though HSI delivers rich data with its continuous spectral bands, it simultaneously presents challenges due to its high dimensionality, potential redundancy, and noise interference. To tackle these challenges, dimension reduction becomes essential in HSI data processing. Our study integrated 11 explainable artificial intelligence (XAI) algorithms with convolutional neural networks (CNNs) to craft an advanced wavelength selection method tailored for honey product analysis. We developed two CNN models and applied five normalization techniques to differentiate honey derived from 23 distinct botanical sources. Through the XAI algorithms, we determined the significance of each spectral wavelength, pinpointing the most characteristic bands. We then used the top eight spectral bands—selected from the averaged significance scores across the XAI algorithms—to optimize the combination of CNN classifiers and normalization techniques. This approach not only streamlines HSI data analysis but also validates the robustness of our XAI-based feature selection methodology. Experimental results demonstrated that the CNN achieved a macro average F1 score of ≥0.99 using just 8 spectral bands selected by averaging the feature importance scores from different XAI algorithms. The generalization ability of our proposed band selection method was further demonstrated by experiments on three HSI remote sensing databases. Overall, our research highlights the potential of XAI algorithms in feature selection, significantly reducing the dimensionality of HSI data without compromising performance. This approach establishes a methodological foundation for designing a cost-effective multispectral imaging system capable of distinguishing the botanical origin of honey.