Convolutional neural networks (CNNs) and training-free CNN variants have been successfully applied to hyperspectral image (HSI) processing and analysis. Training-free CNNs have shown promising feature extraction performance, which could effectively address the issue of typical CNNs being highly parameterized; however, inevitable noise and redundancy in the randomly selected training-free convolutional kernels often leads to unsatisfactory performance. To address this issue, we propose Superpixel Random Selection Random Walk Multi-Branch Depthwise Convolutional Neural Network (SRSRWMD-CNN). Specifically, we propose a novel training-free convolutional neural network characterized by inter-layer multi-scale integration and intra-layer grouping. Various superpixels groups are first generated through multi-scale superpixel segmentation algorithms, then the predetermined number of superpixels are randomly sampled from these groups to serve as training-free convolution kernels. This mechanism enables adaptive computation of HSI feature maps without costly model training in the feature extraction stage, allowing the network to effectively capture a multi-scale spectral-spatial feature representation. Additionally, we propose a multi-branch depthwise convolution strategy that mitigates feature learning errors while significantly enhancing feature representation capabilities. A random walk strategy is employed to expand the receptive field and enhance the robustness of the training-free convolution kernels. Finally, the multi-scale spectral-spatial features are concatenated with the multiple convolutional stages to fuse salient shallow and deep features for accurate HSI classification. Extensive experiments demonstrate that the proposed method achieves superior performance compared to state-of-the-art algorithms.

Tropospheric delay is one of the major error sources in interferometric synthetic aperture radar (InSAR) deformation monitoring. Conventional spatiotemporal filtering tends to oversmooth deformation signals, elevation-dependent phase models are mainly applicable to stratified atmospheric delay, and the direct use of external atmospheric products may introduce additional errors because of spatiotemporal mismatches and data-quality limitations. To address these issues, this article proposes a robust PCA-based atmospheric correction method that integrates multisource external atmospheric data. First, the interferograms are preprocessed to reduce the effects of the flat-Earth term, elevation-dependent phase, and orbital errors. Then, the preprocessed interferometric phase and external atmospheric data, such as the generic atmospheric correction online service (GACOS) and ERA5, are jointly assembled into an observation matrix. A maximum-likelihood robust M-estimator and the median absolute deviation (MAD) are introduced to obtain robust estimates of the variances and covariances of the joint matrix, on which robust principal component analysis (RPCAs) is performed to extract the common atmospheric component while suppressing the influence of outliers, nonshared components, and residual noise. Finally, the robust small-baseline subset (RSBAS) method is incorporated to further refine the time-series results. Simulation experiments and real-data analysis over the Jinsha River Basin demonstrate that the proposed method can effectively mitigate atmospheric delay effects and improve the stability of deformation-rate estimation and time-series inversion, providing an effective solution for high-precision InSAR deformation monitoring in regions with complex atmospheric conditions.

With the development of ion accelerators toward high power, high energy, and high beam intensity, higher requirements have been imposed on the reliability and stability of accelerator power supplies. Variations in ambient temperature directly affect the accuracy and stability of current measurement because of the temperature sensitivity of current sensors and their front-end measurement circuits. Methods relying on component selection or hardware compensation are insufficient to achieve the required current accuracy under wide temperature ranges and complex operating conditions. Based on artificial intelligence optimization techniques, this paper proposes and implements an MLP-based method for modeling and digitally compensating temperature-induced current shift. The models are established based on the relationship between temperature variation and current shift, and a neural-network compensation model incorporating statistical features is further designed. The compensation algorithm is deployed on a domestic FPGA platform to achieve online correction of current measurement data in power supplies. The main novelty of this work lies in the system integration and hardware-algorithm co-design. Experimental results indicate that the proposed method can improve current accuracy and long-term stability, providing an economical and feasible intelligent solution for enhancing the precision of power supplies.

Image inpainting aims to reconstruct missing pixels in images. Recently, generative adversarial networks (GANs) have achieved good performance in image inpainting. However, existing GANs-based methods still perform poorly on large damaged regions because they often require large amounts of training data and tend to lose detail information due to excessive dimensional compression in the generator. To address these problems, we propose an image inpainting method based on multiscale and polarized self-attention generative adversarial network, called MPA-GAN. Multiscale blocks are used to simultaneously learn the detailed information of low-level features and the rich semantic information of high-level features to improve the restoration of large damaged images. In addition, a polarized self-attention module is employed to alleviate information loss caused by dimensional compression, enabling better recovery of both small and large missing areas. Experiments are conducted on three commonly used image inpainting datasets: Paris StreetView, CelebA-HQ, and Places2. The results show the proposed methods achieved good performance on large-area image inpainting.

Lossless compression of remote sensing images is critically important for minimizing storage requirements while preserving the complete integrity of the data. The main challenge in lossless compression lies in striking a good balance between reasonable compression durations and high compression ratios. In this article, we introduce an innovative lossless compression framework that uniquely utilizes lossy compression data as prior knowledge to enhance the compression process. Our framework employs a checkerboard segmentation technique to divides the original remote sensing image into various subimages. The main diagonal subimages are compressed using a traditional lossy method to obtain prior knowledge for facilitating the compression of all subimages. These subimages are then subjected to lossless compression using our newly developed lossy prior probability prediction network (LP3Net) and arithmetic coding in a specific order. The proposed LP3Net is an advanced network architecture, consisting of an image preprocessing module, a channel enhancement module, and a pixel probability transformer module, to learn the discrete probability distribution of each pixel within every subimage, enhancing the accuracy and efficiency of the compression process. Experiments on high-resolution remote sensing image datasets demonstrate the effectiveness and efficiency of the proposed LP3Net and lossless compression framework, achieving a minimum of 4.57% improvement over traditional compression methods and 1.86% improvement over deep learning-based compression methods.

Hyperspectral image (HSI) clustering is challenging to partition pixels into different clusters due to the complex spatial distribution and high-correlated spectrum. Subspace clustering is a representative learning paradigm and has shown competitive performance in HSIs. Most existing methods ignore potential spatial or structural information and show difficulties in dealing with large-scale HSIs. In this paper, we propose an elastic graph fusion subspace clustering (EGFSC) framework that can flexibly incorporate spectral, spatial and structural information for large HSI clustering. Instead of performing pixel-level learning, superpixel-level learning is conducted according to the generated superpixels to lessen computation burden and memory cost. To explore structural information in two perspectives, a superpixel graph and a band graph are constructed based on the superpixel features. Considering the incompatible sizes of the two graphs, we present three effective dual graph fusion strategies to fuse them in different ways. With these graph fusion strategies, EGFSC is able to improve clustering performance by simultaneously considering spatial and structural information. To solve the proposed framework, we present a closed-form solution for easy implementation. Experiments demonstrate that the proposed EGFSC obtains 70.08%, 75.76%, 87.28% and 77.23% clustering accuracies on the four HSI datasets and outperforms the state-of-the-art methods. The source code is released at https://github.com/ZhangYongshan/EGFSC.

Accelerator as a large scale of electrical facilities is trending towards high energy, high intensity and high reliability in the future. Next generation of accelerator will require unprecedented stability of distribution networks. Traditional monitoring methods have some limitations when detecting and analyzing subtle and complex power quality problems. A big data platform built for HIRFL(The Heavy Ion Research Facility in Lanzhou) distribution network integrates various big data technologies and combines historical data with online operational data of the distribution network. Based on this platform, this paper constructs, trains and deploys BP neural network models to identify faults in the distribution network. The results demonstrate that the models have outstanding capabilities of learning and generalization accompanied by accurate identification and classification of faults. This method is significant for the informatization and intelligence of future accelerator distribution networks.

Many tasks, such as image denoising, can be framed within the context of subspace recovery. For its algorithm design, robustness is a critical consideration. In this paper, we propose a novel holistic approach to robust subspace recovery. The fundamental work consists of extending Stein's unbiased risk estimate to elliptical densities, expanding Gaussian scale mixtures, and estimating error density from the dataset. These advancements serve as the foundation for a rescaled three-mode principal component analysis. By leveraging the majorization-minimization (MM) algorithm, we seamlessly integrate total variation into our model. A key feature of this approach is its inherent robustness to outliers, as demonstrated through our experimental results.

Hyperspectral image (HSI) clustering has attracted increasing attention in recent years, because it doesn't rely on labeled pixels. However, it is a challenging task due to the complex spectral-spatial structure. The emergence of large-scale HSIs introduces a new challenge in terms of heightened computational complexity. To address the above challenges, in this paper, we propose a structured anchor projected clustering (SAPC) model for large-scale HSIs. Specifically, we exploit spatial information reflecting in the generated superpixels to perform denoising and generate anchors. Based on the preprocessing, we simultaneously learn a pixel-anchor graph and an anchor-anchor graph in a projected feature space. Meanwhile, the rank-constraint is imposed on the Laplacian matrix related to the anchor-anchor graph. To uncover the clustering structure, we design a clustering inference strategy to propagate clustering labels from anchors to pixels based on the dual graphs. Additionally, we propose an efficient optimization strategy for the formulated SAPC model with linear time complexity in terms of the number of pixels. Since the anchor-anchor graph is with much smaller size, it is high efficient to obtain the structured anchors with pseudo labels. Thus, the clustering process is significantly accelerated. Extensive experiments on multiple large-scale HSI datasets demonstrates the superiority of our SAPC over the state-of-the-art methods. The source code is released at https://github.com/ZhangYongshan/SAPC.

Hyperspectral image (HSI) classification is a key task in the field of remote sensing, aiming to assign category labels to each pixel by leveraging the spectral and spatial information in HSIs. Recently, many deep learning (DL) methods, such as convolutional neural networks (CNNs) and Transformers, have been applied to this task, achieving significant results. However, most existing patch-based DL methods often overlook the potential relationships between the central pixel and its surrounding pixels. Additionally, the unique spectral characteristics of HSIs, such as the high correlation between adjacent spectral bands and low dependence between distant bands, also require special attention. Based on this, we propose a novel dual-branch convolution-Transformer network with spectral-spatial attention (CTSSA), which can effectively aggregate both local and global spectral-spatial features. Specifically, CTSSA comprises two core modules: the Pyramid Spectral Attention Module (PSAM) and the Center Transformer Encoder (CenterTE). The former extracts highly discriminative spectral features through a hierarchical multi-scale attention mechanism, capturing subtle differences between adjacent spectral bands. The latter improves the original Transformer encoder (TE) by introducing a center-attention mechanism to model the global relationship between the central pixel and its surrounding pixels, thereby enhancing classification accuracy while reducing computational complexity. Experimental results on four public datasets (Salinas, Pavia University, Houston, and WHUHi-LongKou) demonstrate that, compared with nine other networks, CTSSA achieves satisfactory performance with fewer parameters and relatively high efficiency.

Pansharpening of multispectral (MS) images aims to merge MS and panchromatic (PAN) images to produce high-resolution multispectral (HRMS) images. Unsupervised pansharpening algorithms, which are widely used in deep learning for pansharpening tasks, commonly employ generative adversarial networks (GANs). However, existing unsupervised methods based on GANs have some limitations: 1) restricted ability for joint preservation of spatial and spectral information and 2) the receptive field of general convolutions restricts the extraction of long-range dependencies. To address these issues, we propose a frequency-guided generative adversarial networks (FG-GANs) for unsupervised pansharpening. Our unsupervised framework uses high-frequency and low-frequency information as prior constraints to guide the training of the FG-GAN's generator and discriminator networks, thereby enhancing the joint preservation of spatial and spectral details. Furthermore, a graph convolution (GCN)-based generator network is designed, in which long-range edge dependencies are extracted and propagated by learning the relationships between distant edge feature nodes. Extensive experiments on the Quickbird and Gaofen-2 datasets demonstrate the effectiveness of our design: our method enhances spatial details and reduces the spatial distortion index (D-s) by 44.8%, while achieving the fidelity of spatial-spectral information with a HQNR score of 0.99.

The hyperspectral anomaly detection (HAD) aims to identify potential anomalies from complex backgrounds. Most reconstruction-based autoencoders equally treat background pixels and anomalies or ignore potential spatial information. In this letter, we propose an HAD method based on multiscale memory autoencoder and spatial filtering, abbreviated as SFM2AE. Specifically, by introducing memory modules into different hidden layers of the autoencoder, multiscale reconstruction of background and anomaly pixels is achieved in the spectral domain. In addition, morphological filtering in the spatial domain is used to extract spatial structural information from anomalies. Joint spatial-spectral anomaly detection is achieved by combining multiscale memory autoencoder and spatial filtering. Experiments demonstrate superior detection performance of the proposed method over the state-of-the-art methods.

Lutan-1 is the first L-band SAR satellite launched by China with the core mission of geohazard monitoring, but few studies have been conducted to apply it in the field of earthquakes. In this paper, the capability of Lutan-1 data in coseismic deformation analysis and seismogenic fault parameter inversion was discussed by taking the 2023 Mw6.0 Jishishan earthquake as an example. Firstly, we utilized Lutan-1 data to acquire the coseismic deformation field of the Jishishan earthquake. Subsequently, the seismogenic fault parameter and slip distribution were inverted using both uniform slip and distributed slip models. Finally, a comprehensive comparison was conducted with Sentinel-1 data in terms of the coseismic deformation field, seismic source parameters, and coherence. The comparative results demonstrate that the coseismic deformation and seismogenic fault parameter inversion derived from Lutan-1 data are consistent with those obtained from Sentinel-1 data. Moreover, Lutan-1 data exhibit superior image quality and better coherence, confirming the effectiveness and superiority of Lutan-1 data for coseismic deformation and seismogenic fault analysis. This study provides a theoretical foundation for the application of Lutan-1 in the field of earthquake disaster monitoring.

Anomaly detection for hyperspectral images (HSIs) is a challenging problem to distinguish a few anomalous pixels from a majority of background pixels. Most existing methods cannot simultaneously explore both structural and spatial information from global and local perspectives. In this letter, we propose a stacked graph fusion denoising autoencoder (SGFDAE) for hyperspectral anomaly detection. Specifically, the global and local graphs are constructed from an HSI to explore potential structural and spatial information. With the designed graph fusion strategy, an advanced graph denoising autoencoder with deep architecture is developed in a hierarchical manner. To achieve better reconstruction and detection, a greedy layerwise unsupervised pretraining strategy is presented for network training. Experiments show that SGFDAE achieves 97.17%, 98.43%, and 98.90% detection accuracies by averaging the results of the datasets from three different scenes and outperforms the state-of-the-art methods.

The upper Jinsha River, located in a high-mountain gorge with complex geological features, is highly prone to large-scale landslides, which could result in the formation of dammed lakes. Analyzing the movement characteristics of the typical Xiaomojiu landslide in this area contributes to a better understanding of the dynamics of landslides in the region, which is of great significance for landslide risk prediction and analysis. True displacement data on the surface of landslides are crucial for understanding the morphological changes in landslides, providing fundamental parameters for dynamic analysis and risk assessment. This study proposes a method for calculating the actual deformation of landslide bodies based on multi-track Interferometric Synthetic Aperture Radar (InSAR) deformation data. It iteratively solves for the optimal true deformation vector of the landslide on a per-pixel basis under a least-squares constraint based on the assumption of consistent displacement direction among adjacent points on the landslide surface. Using multi-track Sentinel data from 2017 to 2023, the line of sight (LOS) accumulative de-formation of the Xiaomojiu landslide was obtained, with a maximum LOS deformation of −126 mm/year. The true surface displacement of the Xiaomojiu landslide after activation was calculated using LOS deformation. The development of two rotational sub-slipping zones on the landslide body is inferred based on the distribution of actual displacements along the central profile line. Analysis of temporal changes in water body area data revealed that the Xiaomojiu landslide was activated after a barrier lake event and continuously moved due to the influence of higher water levels’ in the river channel. In conclusion, the proposed method can be applied to calculate the true surface displacement of landslides with complex mechanisms for analyzing the movement status of landslide bodies. Furthermore, the spatiotemporal analysis of the Xiaomojiu landslide characteristics can support analyzing the mechanisms of similar landslides in the Jinsha River Basin.

Data Augmentation (DA) is significant for Hyperspectral Image (HSI) classification especially in the case of limited labeled training data. Various DA models have been introduced to augment HSI data by using image processing techniques, prior knowledge or contextual information of HSI. However, spatial information in HSI could be inefficiently adopted in these DA models, leading to the lack of diversity and richness of the augmented data. To handle the issue, inspired by the recently proposed Principal Component Analysis based DA (PCA-DA) and Superpixelwise PCA (SuperPCA), we introduce SuperPCA based DA (SuperPCA-DA) in this paper. Specifically, an HSI is firstly divided into various superpixel blocks by typical image segmentation techniques, followed by fixed-size window and superpixel based local reconstruction for HSI denoising, and then the proposed SuperPCA-DA is employed in each superpixel block for HSI data augmentation where PCA is locally conducted in each superpixel to extract low-dimensional features, which will be projected back onto the original high-dimensional spectral space with random noises added into the projection matrix of PCA. The novel DA model can effectively generate new samples with diversity and richness by employing superpixels based spatial information of HSI, which could outperform classic and state-of-the-art DA models and then improve the accuracy of the subsequent classification models, especially when the number of training data is limited. The experimental results on three HSI datasets demonstrate the effectiveness of the proposed method. The code of the proposed model is available at https://github.com/XinweiJiang/SuperPCA-DA .

Multimodal remote sensing image recognition aims to identify a category of land cover for every pixel with consistency and complementary information provided by different modalities. Most existing methods perform land cover recognition in a supervised manner with explicit label guidance. It is challenging to perform recognition without label guidance due to the complex spatial distribution and modality incompatibility, especially for large-scale data. In this article, we propose a dual graph learning affinity propagation (DGLAP) method for multimodal remote sensing image clustering. Based on the consistent spatial distribution from local regions, the proposed method learns an N x M consensus anchor graph from N denoised pixels and M anchors by adaptive weighting different modalities along with projection learning. Meanwhile, an optimal M x M compressed consensus anchor graph is learned from the updated anchors in different modalities with diverse adaptive contributions and connectivity constraint. Since M << N , clustering results can be efficiently obtained according to affinity propagation from the pseudolabeled anchors to the pixels without additional steps. An alternating optimization algorithm is devised to solve the proposed formulation. This is the first attempt to propose a ultraefficient graph-based clustering method with linear time complexity O(N) and low time cost for large-scale multimodal remote sensing data. Extensive experiments on three datasets demonstrate the superiority of the proposed method over the state-of-the-art methods in both efficacy and efficiency. The code is released at https://github.com/ZhangYongshan/DGLAP .

The big data platform for accelerator distribution network deals with a large amount of data every moment during operation. Collecting and analyzing these data to identify hidden causal relationships within system makes it possible to effectively locate the root cause of system faults and improve system maintenance efficiency. Applying the methods of Kernel Principal Component Analysis (KPCA) and Granger Cointegratance (GC) to this platform, we first collect relevant information on power quality of accelerator distribution network under different faults conditions in order to understand changes between information before and after fault. Then, the KPCA method is used to extract the features of a large amount of fault data. Finally, the GC algorithm is utilized to identify the causal relationships between data from different locations. This paper would provide a effectual method for analysis, discovery and optimization of causes of faults in the accelerator distribution network.

Dimensionality reduction (DR) is important for feature extraction and classification of hyperspectral images (HSIs). Recently proposed superpixel-based DR models have shown promising performance, where superpixel segmentation techniques were applied to segment an HSI and then DR models like principal component analysis (PCA) or linear discriminant analysis (LDA) were employed to extract the local and/or global features. However, superpixelwise PCA (SuperPCA)-based local features are unsatisfactory because PCA aims to extract features with high variance, which could be inefficient in superpixels with mixed objects or strong noise/outliers. In addition, superpixelwise unsupervised LDA (SuperULDA) based global features may neglect local (spatial-contextual) information. To address these issues, we propose a new spectral–spatial and superpixelwise unsupervised LDA (S3-ULDA) model for unsupervised feature extraction from HSIs. Specifically, the HSI is first segmented into various superpixels with pseudo labels. Then, superpixel-based local reconstruction for HSI denoising is conducted. Next, SuperULDA is performed on both the original HSI and locally reconstructed data to extract global features. Then, superpixelwise unsupervised local Fisher discriminant analysis (SuperULFDA) is developed for local feature extraction, where each superpixel and its adjacent superpixels (along with their pseudo-labels) are fed into local Fisher discriminant analysis (LFDA) to extract local features. The superpixel-level local manifold structures can be effectively modeled by the proposed SuperULFDA. Finally, by fusing the extracted global and local features, novel global–local and spectral–spatial features can be obtained. Our experimental results on several benchmark HSIs demonstrate the superiority of the proposed method over state-of-the-art methods. The code of the proposed model is available at https://github.com/XinweiJiang/S3-ULDA.

Predictions of oil and gas reservoirs play an essential role in hydrocarbon reservoir exploration. Multi-attribute seismic data can provide abundant reservoir information. However, traditional reservoir prediction models only reflect the characteristics of local seismic channels, ignoring information on global geological spatial structures. Thus, a semi-supervised prediction model based on self-attention is proposed to extract reservoirs' oil and gas characteristics. With the high cost of labeled samples, neighborhood similarity measurement based on Grey Relation Analysis is applied to determine the homogeneity of adjacent seismic traces for the first time to obtain pseudo-labels while considering geological spatial structure. As a semi-supervised prediction model, a multilayer convolutional neural network with a multi-head self-attention mechanism is utilized, which can extract features from multiple dimensions and investigate global structural information. The experimental results demonstrate that the algorithm in this paper's reservoir oil and gas prediction is more accurate than that of conventional approaches and is consistent with real drilling data.