Hyperspectral Data Sets Research Articles

SSFN: a novel Spatial-Spectral FusionNet for hyperspectral image classification

Hyperspectral image classification is the process of identifying ground objects within hyperspectral images at the pixel level. While many CNN based methods have been successfully used in this field, they often struggle to effectively extract inter-spectral information due to the high spectral dimensions of hyperspectral images. Recently, the transformer model has been introduced to capture the interdependence between spectral data, but it may lose the ability to capture local context features. In order to address the limitations of CNN and Transformer models, we propose a novel approach that integrates the advantages of both models. Specifically, we use CNN to extract spatial information and Transformer to extract spectral information, which is then fused before being fed into the MLP framework for classification. Additionally, we introduce a sparse strategy to eliminate the impact of redundant frequency bands on the transformer’s performance. Our method fully utilizes the spatial and spectral information in hyperspectral image data and has achieved excellent performance on hyperspectral datasets. By leveraging both CNN and transformer models, we can effectively capture both spatial and spectral features, providing a powerful tool for hyperspectral image classification.

Hyperspectral remote sensing image classification based on residual generative Adversarial Neural Networks

Classification of hyperspectral images is an essential application of deep learning techniques. However, standard deep learning approaches require a large number of labelled samples for model training, and classification performance can be enhanced. In this study, we propose the use of residual generative adversarial networks for the classification of hyperspectral images based on deep learning Convolutional Neural Networks, a large number of labelled samples, and the high classification accuracy required for the classification task. The method is based on generative adversarial networks, with a 6-layer residual network containing up-sampling and convolutional layers replacing the inverse convolutional layer network structure of the generator to enhance the data generation capability, and an 18-layer residual convolutional network replacing the convolutional layer network structure of the discriminator to enhance the feature extraction capability. The hyperspectral image classification method with residual generation adversarial network enhances the information exchange between the shallow network and the deep network by adding a residual structure to the network, extracts the deep features of hyperspectral images and improves the accuracy of hyperspectral image classification. Extensive experiments on several benchmark hyperspectral datasets have shown that the method outperforms comparative methods by 0.7% to 22.3% on OA.

Hyperspectral image classification using improved multi-scale block local binary pattern and bi-exponential edge-preserving smoother

ABSTRACT In this paper, a multi-strategy fusion (MSF) framework, based on improved MBLBP and bi-exponential edge-preserving smoother (BEEPS), is proposed for hyperspectral image (HSI) classification. First, MBLBP operator is adopted to characterize the overall structural information of HSI, where the averaging strategy allocates same weights for the pixels in a local sub-region, so that the edges tend to be blurred due to being isotropic. To solve this question, the steering kernel is first introduced into MBLBP for learning the local structure prior of HSI. Then, a support vector machine classifier is used to calculate the soft classified probabilities of pixels. Furthermore, BEEPS is adopted to smooth the soft classified probabilities maps in the post-processing stage, and the purpose is to further improve classification accuracy of HSI by considering context-aware information for each class label. Experiments are performed on three real hyperspectral datasets, namely, Indian Pines, KSC, and Houston 2013, only 1%, 6, and 5 labeled samples are randomly selected for training, the overall accuracy(kappa) obtained by MSF is 99.47%(99.40), 99.52%(99.47), and 94.25%(93.78), respectively, which is far better than the contrast methods.

Spectral Swin Transformer Network for Hyperspectral Image Classification

Hyperspectral images are complex images that contain more spectral dimension information than ordinary images. An increasing number of HSI classification methods are using deep learning techniques to process three-dimensional data. The Vision Transformer model is gradually occupying an important position in the field of computer vision and is being used to replace the CNN structure of the network. However, it is still in the preliminary research stage in the field of HSI. In this paper, we propose using a spectral Swin Transformer network for HSI classification, providing a new approach for the HSI field. The Swin Transformer uses group attention to enhance feature representation, and the sliding window attention calculation can take into account the contextual information of different windows, which can retain the global features of HSI and improve classification results. In our experiments, we evaluated our proposed approach on several public hyperspectral datasets and compared it with several methods. The experimental results demonstrate that our proposed model achieved test accuracies of 97.46%, 99.7%, and 99.8% on the IP, SA, and PU public HSI datasets, respectively, when using the AdamW optimizer. Our approach also shows good generalization ability when applied to new datasets. Overall, our proposed approach represents a promising direction for hyperspectral image classification using deep learning techniques.

DSSFN: A Dual-Stream Self-Attention Fusion Network for Effective Hyperspectral Image Classification

Hyperspectral images possess a continuous and analogous spectral nature, enabling the classification of distinctive information by analyzing the subtle variations between adjacent spectra. Meanwhile, a hyperspectral dataset includes redundant and noisy information in addition to larger dimensions, which is the primary barrier preventing its use for land cover categorization. Despite the excellent feature extraction capability exhibited by convolutional neural networks, its efficacy is restricted by the constrained receptive field and the inability to acquire long-range features due to the limited size of the convolutional kernels. We construct a dual-stream self-attention fusion network (DSSFN) that combines spectral and spatial information in order to achieve the deep mining of global information via a self-attention mechanism. In addition, dimensionality reduction is required to reduce redundant data and eliminate noisy bands, hence enhancing the performance of hyperspectral classification. A unique band selection algorithm is proposed in this study. This algorithm, which is based on a sliding window grouped normalized matching filter for nearby bands (SWGMF), can minimize the dimensionality of the data while preserving the corresponding spectral information. Comprehensive experiments are carried out on four well-known hyperspectral datasets, where the proposed DSSFN achieves higher classification results in terms of overall accuracy (OA), average accuracy (AA), and kappa than previous approaches. A variety of trials verify the superiority and huge potential of DSSFN.

Advancing peatland vegetation mapping by spaceborne imaging spectroscopy

Peatlands contribute to a wide range of ecosystem services. They play an important role as carbon sinks in their natural state, but when they are drained, they causecarbon emissions. Rewetting drained peatlands is required to reduce carbon emissions and create new carbon sinks. However, drained peatlands are commonly used as grassland or croplands; therefore, alternative agriculture schemes are required following rewetting. Paludiculture, i.e., agriculture on wet and rewetted peatlands, is an option in these areas after rewetting to produce biomass sustainably. Monitoring of peatland management is challenging, yet needed to ensure a successful rewetting and plantation of, e.g., Phragmites australis and Typha spp., two plants which are commonly used in paludiculture. Remote sensing is an excellent tool for monitoring the vegetation composition of vast rewetted peatland regions. However, because many peatland species have similar spectral characteristics, such monitoring is ideally based on high-spatial, high-temporal hyperspectral images. Data that complies with all these requirements does not exist on a regular basis. Therefore, we assessed the potential for mapping peatland vegetation communities in the Peene and Trebel river basins of the federal state of Mecklenburg-Western Pomerania, Germany, using multi-date hyperspectral (PRISMA) data. We used regression-based unmixing to map fractions of different peatland vegetation classes. Results were analyzed with regard to the contribution of multi-date observations and, in comparison, to multispectral datasets (Landsat-8/Sentinel-2). Our results showed that different classes are best mapped at different observation dates. The multi-date hyperspectral datasets produced less Mean Absolute Error (MAE = 16.4%) than the single-date hyperspectral images (ΔMAE + 1%), with high accuracies for all classes of interest. Compared to the results obtained with multispectral data from similar acquisition dates and annual spectral-temporal metrics (STM), the results from hyperspectral data were always clearly superior (ΔMAE + 4%). Besides the superior performance during comparisons, our results also indicate that information that can be derived from the hyperspectral data with the regression-based unmixing goes clearly beyond that of discrete classification. With more hyperspectral sensors coming up and an expected higher availability of multi-data hyperspectral imagery, these data can be expected to play a bigger role in the future monitoring of peatlands.

Diagnosis and correction methods for spectral interference in the framework of LIBS imaging

Laser-Induced Breakdown Spectroscopy (LIBS) has become a powerful imaging technique for elemental characterization in analytical chemistry due to its advantages over other techniques. Major, minor, and trace elements are detected with high measurement dynamic, a low limit of detection and a high acquisition rate, allowing for the quick analysis of large sample surfaces. Today, chemometric tools are commonly used to ensure the most comprehensive and unbiased exploration of such spectroscopic data. However, the integration of the signal from a wavelength assumed to be specific to the element of interest remains the basic tool for generating a chemical distribution map from a hyperspectral dataset. This classical approach is based on a strong assumption, the specificity of the chemical information on the spectral domain being considered. Any spectral interference inevitably result in the generation of a biased distribution image. In this publication, we demonstrate how Principal Component Analysis (PCA) can diagnose the potential presence of a spectral interference and how Multivariate Curve Resolution-Alternating Least Squares (MCR-ALS) can ultimately correct it if necessary using a LIBS imaging dataset obtained from the analysis of a complex rock sample. The proposed approach combines the simplicity and effectiveness of the integration method with the diagnostic and correction capabilities of chemometric tools, providing a comprehensive solution for spectral interference in LIBS imaging.

Unsupervised dimensionality reduction of medical hyperspectral imagery in tensor space

Background and objectivesCompared with traditional RGB images, medical hyperspectral imagery (HSI) has numerous continuous narrow spectral bands, which can provide rich information for cancer diagnosis. However, the abundant spectral bands also contain a large amount of redundancy information and increase computational complexity. Thus, dimensionality reduction (DR) is essential in HSI analysis. All vector-based DR methods ignore the cubic nature of HSI resulting from vectorization. To overcome the disadvantage of vector-based DR methods, tensor-based techniques have been developed by employing multi-linear algebra. MethodsTo fully exploit the structure features of medical HSI and enhance computational efficiency, a novel method called unsupervised dimensionality reduction via tensor-based low-rank collaborative graph embedding (TLCGE) is proposed. TLCGE introduces entropy rate superpixel (ERS) segmentation algorithm to generate superpixels. Then, a low-rank collaborative graph weight matrix is constructed on each superpixel, greatly improving the efficiency and robustness of the proposed method. After that, TLCGE reduces dimensions in tensor space to well preserve intrinsic structure of HSI. ResultsThe proposed TLCGE is tested on cholangiocarcinoma microscopic hyperspectral data sets. To further demonstrate the effectiveness of the proposed algorithm, other machine learning DR methods are used for comparison. Experimental results on cholangiocarcinoma microscopic hyperspectral data sets validate the effectiveness of the proposed TLCGE. ConclusionsThe proposed TLCGE is a tensor-based DR method, which can maintain the intrinsic 3-D data structure of medical HSI. By imposing the low-rank and sparse constraints on the objective function, the proposed TLCGE can fully explore the local and global structures within each superpixel. The computational efficiency of the proposed TLCGE is better than other tensor-based DR methods, which can be used as a preprocessing step in real medical HSI classification or segmentation.

The use of a multi-sensor core scanner workflow as the backbone of a digital core repository

Abstract A core repository is a physical record of a country's or commercial organization's subsurface wealth. Some of the largest core repositories hold thousands of kilometres of core material and it is a challenge to turn this physical archive into an accessible digital resource for all. Non-destructive multi-sensor core logger, hyperspectral and X-ray imaging techniques offer a unique chance to rescue valuable data trapped within core samples, improving the way that a core repository delivers data to academic or industrial end users. Here we present a case study of an archived petroleum core acquired in 1985 at the Osprey Field, UK Continental Shelf. Data from the UK National Data Repository are augmented by a multi-sensor core logger, hyperspectral and X-ray dataset that is uploaded into a cloud-based digital repository. The data were analysed using a multi-variant analysis to reclassify the original lithological interpretations, uncovering a greater proportion of clay and cemented horizons than was previously interpreted. A workflow is established to optimize the use of legacy cores and exploit the abundance of data trapped within the core repository using continuous multi-sensor core scanning and imaging data, which are stored within the virtual environment for visualization and access to all.

Improved SR-SSIM Band Selection Method Based on Band Subspace Partition

Scholars have performed much research on reducing the redundancy of hyperspectral data. As a measure of the similarity between hyperspectral bands, structural similarity is used in band selection methods. However, existing structural similarity methods calculate all the structural similarity between bands, which leads to excessively long runtimes for these methods. Aiming to address this problem, this paper proposes a band subspace partition method and combines it with the SR-SSIM band selection method to obtain an improved band selection method: E-SR-SSIM. E-SR-SSIM consists of two parts: band subspace partition and band subspace band selection. In the first part, the hyperspectral dataset is divided into subdatasets corresponding to a number of subspaces. In the second part, a modified SR-SSIM method is used for all subdatasets to select the most representative band in each subdataset. The Indian Pines, Salinas Kennedy Space Center and Wuhan unmanned aerial vehicle-borne hyperspectral image LongKou public datasets are used to implement the experiment. The experiment uses random forest as the supervised classifier: 10% of each category sample is randomly selected as training data, and the remaining 90% is used as test data. The evaluation indicators selected in the experiment are overall accuracy, average accuracy, kappa and recall. The experimental results show that E-SR-SSIM can effectively reduce the runtime while ensuring classification compared with SR-SSIM, and quantitative proof that the band subspace partition reduces the calculated amount of structural similarity is obtained through a mathematical analysis. The improved band subspace partition method could partition a dataset more reasonably than the original band subspace partition method.

Estimation of leaf photosynthetic capacity parameters using spectral indices developed from fractional-order derivatives

The maximum carboxylation rate (Vcmax) and the maximum electron transport rate (Jmax) are representative parameters describing the gas exchange properties of leaves. Non-destructive measurements of these parameters are valuable for tracing the physiological status and productivity of a plant. The spectral vegetation index, as a simple and efficient approach, has been applied to estimate these two parameters. However, most indices used are derived from original or integer-order rather than fractional-order derivatives (FOD), despite increasing evidence of their advantages in terms of tracing diverse plant properties. In this study, we attempt to derive photosynthetic capacity parameters from FOD indices based on a composite hyperspectral spectroscopy dataset covering various conditions. Our results show that the FOD can be utilized to develop more robust indices for obtaining photosynthetic capacity parameters. In particular, the D-type indices calculated from the 1.50-order and 1.60-order derivatives produced the highest accuracies for the estimation of Vcmax and Jmax. The performance of these proposed indices, however, differed for different plant functional types and had a high sensitivity to random noise. The results of the present research show that FOD indices have the potential to estimate leaf photosynthetic parameters, and are therefore to support hyperspectral information on estimating plant properties.

A Deep Learning-Based Hyperspectral Object Classification Approach via Imbalanced Training Samples Handling

Object classification in hyperspectral images involves accurately categorizing objects based on their spectral characteristics. However, the high dimensionality of hyperspectral data and class imbalance pose significant challenges to object classification performance. To address these challenges, we propose a framework that incorporates dimensionality reduction and re-sampling as preprocessing steps for a deep learning model. Our framework employs a novel subgroup-based dimensionality reduction technique to extract and select the most informative features with minimal redundancy. Additionally, the data are resampled to achieve class balance across all categories. The reduced and balanced data are then processed through a hybrid CNN model, which combines a 3D learning block and a 2D learning block to extract spectral–spatial features and achieve satisfactory classification accuracy. By adopting this hybrid approach, we simplify the model while improving performance in the presence of noise and limited sample size. We evaluated our proposed model on the Salinas scene, Pavia University, and Kennedy Space Center benchmark hyperspectral datasets, comparing it to state-of-the-art methods. Our object classification technique achieves highly promising results, with overall accuracies of 99.98%, 99.94%, and 99.46% on the three datasets, respectively. This proposed approach offers a compelling solution to overcome the challenges of high dimensionality and class imbalance in hyperspectral object classification.

HyperSFormer: A Transformer-Based End-to-End Hyperspectral Image Classification Method for Crop Classification

Crop classification of large-scale agricultural land is crucial for crop monitoring and yield estimation. Hyperspectral image classification has proven to be an effective method for this task. Most current popular hyperspectral image classification methods are based on image classification, specifically on convolutional neural networks (CNNs) and recurrent neural networks (RNNs). In contrast, this paper focuses on methods based on semantic segmentation and proposes a new transformer-based approach called HyperSFormer for crop hyperspectral image classification. The key enhancement of the proposed method is the replacement of the encoder in SegFormer with an improved Swin Transformer while keeping the SegFormer decoder. The entire model adopts a simple and uniform transformer architecture. Additionally, the paper introduces the hyper patch embedding (HPE) module to extract spectral and local spatial information from the hyperspectral images, which enhances the effectiveness of the features used as input for the model. To ensure detailed model processing and achieve end-to-end hyperspectral image classification, the transpose padding upsample (TPU) module is proposed for the model’s output. In order to address the problem of insufficient and imbalanced samples in hyperspectral image classification, the paper designs an adaptive min log sampling (AMLS) strategy and a loss function that incorporates dice loss and focal loss to assist model training. Experimental results using three public hyperspectral image datasets demonstrate the strong performance of HyperSFormer, particularly in the presence of imbalanced sample data, complex negative samples, and mixed sample classes. HyperSFormer outperforms state-of-the-art methods, including fast patch-free global learning (FPGA), a spectral–spatial-dependent global learning framework (SSDGL), and SegFormer, by at least 2.7% in the mean intersection over union (mIoU). It also improves the overall accuracy and average accuracy values by at least 0.9% and 0.3%, respectively, and the kappa coefficient by at least 0.011. Furthermore, ablation experiments were conducted to determine the optimal hyperparameter and loss function settings for the proposed method, validating the rationality of these settings and the fusion loss function.

Discriminative feature constraints via supervised contrastive learning for few-shot forest tree species classification using airborne hyperspectral images

In scenarios where sample collection is limited, studying few-shot learning algorithms such as prototypical networks (P-Net) is a keynote topic for supervised multiple tree species classification. In a previous study, we improved the P-Net by combining the feature enhancement algorithm based on the convolutional block attention module and several popular data augmentation methods in the computer vision domain, the classification accuracy can be significantly increased, and the degree of model overfitting can be reduced. However, there is a clear boundary between the data augmentations and the feature enhancement algorithm, which is manifested in that data augmentations are only used to enrich the diversity of the learned samples, but cannot directly affect the construction of the objective function, thus limiting the ability of data augmentations. In fact, in the supervised contrastive learning research, data augmentation methods are often used to generate positive samples of an anchor image to construct the objective function, i.e. supervised contrastive loss. The core idea for solving such a boundary problem is to use contrastive learning to make the anchor image close to its positive samples and the negative samples away from each other. Inspired by this, we introduced supervised contrastive learning in the P-Net, i.e., SCL-P-Net, which takes the discriminative feature representations as the constraints of the prototype clustering algorithm. In SCL-P-Net, data augmentation methods can not only extend the sample distribution, but also be used to construct the supervised contrastive loss directly. The study involves four airborne hyperspectral image datasets related to tree species classification, including the GFF-A and GFF-B datasets collected from Gaofeng Forest Farm in Nanning City, Guangxi Province, South China, the Xiongan dataset from Matiwan Village in Xiongan New Area, Hebei Province, North China, and the Tea Farm dataset from Fanglu Tea Farm in Changzhou City, Jiangsu Province, East China. The highest overall accuracy (OA) for the four datasets is 99.23% for GFF-A, 98.39% for GFF-B, 99.30% for Xiongan, and 99.54% for Tea Farm. It is concluded that the proposed SCL-P-Net classification framework can achieve multiple tree species classification with high-precision. Without changing the basic classification framework of P-Net, the introduction of supervised contrastive learning makes the combination of the data augmentations and the feature enhancement algorithm and plays a positive role in improving the distinguishability between samples.

Hyperspectral-to-image transform and CNN transfer learning enhancing soybean LCC estimation

Leaf chlorophyll content (LCC) is a distinct indicator of crop health status used to estimate nutritional stress, diseases, and pests. Thus, accurate LCC information can assist in the monitoring of crop growth. The combined use of hyperspectral and deep learning techniques (e.g., convolutional neural network [CNN] and transfer learning [TL]) can improve the performance of crop LCC estimation. We propose a hyperspectral-to-image transform (HIT) technique for converting canopy hyperspectral reflectance into 2D images. We designed a CNN architecture called LCCNet that fuses the deep and shallow features of CNNs to improve soybean LCC estimation. This study evaluated the combined use of hyperspectral remote sensing (RS), HIT, CNN, and TL techniques to estimate soybean LCC for multiple growth stages. The LCCNet was pre-trained based on a simulated dataset (n = 114,048) from the PROSAIL radiative transfer model (RTM) and used as prior knowledge for this work. The soybean canopy hyperspectral RS dataset (n = 910) was obtained using a FieldSpec 3 spectrometer. The knowledge gained while learning to estimate LCC from PROSAIL RTM was applied when estimating field soybean LCC (Dualex readings). TL was used to enhance the soybean estimation model, called the Soybean-LCCNet (RTM + HIT + CNN + TL) model. We tested the LCC (Dualex readings) estimation performance using (a) HIT + CNN, (b) LCCNet (RTM + HIT + CNN), (c) Soybean-LCCNet (RTM + HIT + CNN + TL), and (d) widely used LCC spectral features + partial least squares regression (PLSR). Four methods were ranked based on their LCC estimation performance: Soybean-LCCNet (R2 = 0.78, RMSE = 4.13 (Dualex readings)) > HIT + CNN (R2 = 0.75, RMSE = 4.41 (Dualex readings)) > PLSR-based method (R2 = 0.61, RMSE = 5.39 (Dualex readings)) > LCCNet (R2 = 0.53, RMSE = 7.11 (Dualex readings)). The main conclusions of this work are as follows: (1) HIT + CNN can provide a more robust LCC estimation performance than the widely used LCC SIs; (2) Fusing the deep and shallow features of CNNs can improve the performance of RS soybean LCC (Dualex readings) estimation; and (3) Soybean-LCCNet can reuse the CNN layer information of a pre-trained LCCNet based on a PROSAIL RTM dataset and improve the soybean LCC estimation performance.

Mapping several soil types using hyperspectral datasets and advanced machine learning methods

Specifying surface soil types is vital for healthy agricultural management to enhance food production. Recent advancements in machine learning are essential in soil science, quantitatively predicting and classifying soil types. The current study concentrates on generating and testing the spectral library to classify surface soil types. It deals with advanced machine learning methods in evaluating soil physicochemical properties and soil type classification using Spectroradiometer and satellite imagery. The proposed methodology tests Phulambri Tehsil's agricultural regions in Aurangabad district, Maharashtra, India. The soil properties determination has been accomplished using partial least square regression (PLSR) models, enabling the relationship between soil profiles and reflectance spectra. The soil type classification has been done on identified values of specific properties in sampled profiles. Machine learning methods such as PLSR, support vector machine (SVM), and spectral angle mapper (SAM), along with principal component analysis (PCA) and minimum-noise-fraction (MNF), were used. The study showed that the three major soil classes have accurately identified and mapped from satellite images using machine learning methods with more than 95% classification accuracy. The results proved to be useful for soil study in heterogeneous areas. Therefore, this study is used in precision farming to enhance food production and management.

A Hyperspectral Space of Skin Tones for Inverse Rendering of Biophysical Skin Properties

AbstractWe present a method for estimating the main properties of human skin, leveraging a hyperspectral dataset of skin tones synthetically generated through a biophysical layered skin model and Monte Carlo light transport simulations. Our approach learns the mapping between the skin parameters and diffuse skin reflectance in such space through an encoder‐decoder network. We assess the performance of RGB and spectral reflectance up to 1 μm, allowing the model to retrieve visible and near‐infrared. Instead of restricting the parameters to values in the ranges reported in medical literature, we allow the model to exceed such ranges to gain expressiveness to recover outliers like beard, eyebrows, rushes and other imperfections. The continuity of our albedo space allows to recover smooth textures of skin properties, enabling reflectance manipulations by meaningful edits of the skin properties. The space is robust under different illumination conditions, and presents high spectral similarity with the current largest datasets of spectral measurements of real human skin while expanding its gamut.

Precise crop classification of UAV hyperspectral imagery using kernel tensor slice sparse coding based classifier

Precise crop classification plays a significant role in the agriculture field. An appropriate data source for precise crop classification is high spatial resolutions hyperspectral imagery (H2 imagery) acquired by unmanned aerial vehicle (UAV). However, for imagery with many different classes of crops, crop classification of UAV H2 imagery is a huge challenge. The significant spectral diversity, spatial heterogeneity and nonlinear data structure of UAV H2 imagery results in poor spectral discriminability. To improve the discriminability, a kernel tensor slice sparse coding-based classifier (KTSSCC) is proposed for precise crop classification of UAV H2 imagery in this research. The kernel tensor representation mechanism in KTSSCC can reduce the nonlinear separation while well preserving the spectral characteristics and spatial constraints of land-covers, and thus the discriminability is greatly improved. Furthermore, this paper puts forward the kernel tensor slice sparse orthogonal matching pursuit (KTSSOMP) algorithm to optimize kernel tensor slice sparse coding in the spectral space, which greatly reduces the computation cost. Moreover, there are very few parameters to be tuned in our proposed model. We assess the performance of KTSSCC on two real UAV hyperspectral imagery datasets, and find that, based on visual and quantitative results, it provides satisfactory crop classification results and outperforms the state-of-the-art approaches.

Shallow-Guided Transformer for Semantic Segmentation of Hyperspectral Remote Sensing Imagery

Convolutional neural networks (CNNs) have achieved great progress in the classification of surface objects with hyperspectral data, but due to the limitations of convolutional operations, CNNs cannot effectively interact with contextual information. Transformer succeeds in solving this problem, and thus has been widely used to classify hyperspectral surface objects in recent years. However, the huge computational load of Transformer poses a challenge in hyperspectral semantic segmentation tasks. In addition, the use of single Transformer discards the local correlation, making it ineffective for remote sensing tasks with small datasets. Therefore, we propose a new Transformer layered architecture that combines Transformer with CNN, adopts a feature dimensionality reduction module and a Transformer-style CNN module to extract shallow features and construct texture constraints, and employs the original Transformer Encoder to extract deep features. Furthermore, we also designed a simple Decoder to process shallow spatial detail information and deep semantic features separately. Experimental results based on three publicly available hyperspectral datasets show that our proposed method has significant advantages compared with other traditional CNN, Transformer-type models.

CRNN: Collaborative Representation Neural Networks for Hyperspectral Anomaly Detection

Hyperspectral anomaly detection aims to separate anomalies and backgrounds without prior knowledge. The collaborative representation (CR)-based hyperspectral anomaly detection methods have gained significant interest and development because of their interpretability and high detection rate. However, the traditional CR presents a low utilization rate for deep latent features in hyperspectral images, making the dictionary construction and the optimization of weight matrix sub-optimal. Due to the excellent capacity of neural networks for generation, we formulate the deep learning-based method into CR optimization in both global and local streams, and propose a novel hyperspectral anomaly detection method based on collaborative representation neural networks (CRNN) in this paper. In order to gain a complete background dictionary and avoid the pollution of anomalies, the global dictionary is collected in the global stream by optimizing the dictionary atom loss, while the local background dictionary is obtained by using a sliding dual window. Based on the two dictionaries, our two-stream networks are trained to learn the global and local representation of hyperspectral data by optimizing the objective function of CR. The detection result is calculated by the fusion of residual maps of original and represented data in the two streams. In addition, an autoencoder is introduced to obtain the hidden feature considered as the dense expression of the original hyperspectral image, and a feature extraction network is concerned to further learn the comprehensive features. Compared with the shallow learning CR, the proposed CRNN learns the dictionary and the representation weight matrix in neural networks to increase the detection performance, and the fixed network parameters instead of the complex matrix operations in traditional CR bring a high inference efficiency. The experiments on six public hyperspectral datasets prove that our proposed CRNN presents the state-of-the-art performance.