The fusion of multisource remote sensing (RS) data has demonstrated significant potential in target recognition and classification tasks. However, there is limited emphasis on capturing both high- low-frequency information from these sources. Additionally, effectively integrating remains a challenging task, as the absence redundancy discriminant hampers applications RS data. In this paper, we propose network called combining global local features (NCGLF2) that integrates (GLF) extracted This...

In complex scenes, dynamic background, illumination variation, and shadow are important factors, which make conventional moving object detection algorithms suffer from poor performance. To solve this problem, a method via ResNet-18 with encoder-decoder structure is proposed to segment objects scenes. possesses pixel-level classification capability divide pixels into foreground it performs well in feature extraction because of its layers so shallow that many more low-scale features will be...

The joint sparse representation (JSR)-based classifier assumes that pixels in a local window can be jointly and sparsely represented by dictionary constructed the training samples. class label of each pixel decided according to residual. However, once includes from different classes, performance JSR may seriously decreased. Since correlation coefficient (CC) is able measure spectral similarity among efficiently, this letter proposes new classification method via fusing CC JSR, which attempts...

Superpixel segmentation has been demonstrated to be a powerful tool in hyperspectral image (HSI) classification. Each superpixel region can regarded as homogeneous region, which is composed of series spatial neighboring pixels. However, may contain the pixels from different classes. To further explore optimal representations superpixels, new framework based on two k selection rules proposed find most representative training and test samples. The method consists following four steps: first,...

The "noisy label" problem is one of the major challenges in hyperspectral image (HSI) classification. In order to address this problem, a spatial density peak (SDP) clustering-based method proposed detect mislabeled samples training set. Specifically, methods consist following steps: first, correlation coefficients among each class are estimated. step, instead measuring by considering individual samples, all neighbor or K representative local window surrounding sample considered. By way,...

Classification is an important technique for remotely sensed hyperspectral image (HSI) exploitation. Often, the presence of wrong (noisy) labels presents a drawback accurate supervised classification. In this article, we introduce new framework noisy label detection that combines superpixel-to-pixel weighting distance (SPWD) and density peak clustering. The proposed method able to accurately detect remove in training set before HSI It considers two weak assumptions when exploiting...

Change detection approaches can detect changed areas of the same scene at different times. Hyperspectral remote-sensing images contain large amounts spectral information high resolution. As hyperspectral datasets become abundant, more and change technologies use as raw data. suffer from band redundancy. There is an urgent need to improve directionality features. To solve these problems, in this article, we propose a CNN framework involving slow-fast selection (SFBS) feature fusion grouping...

Convolutional neural networks (CNNs) have been widely used in hyperspectral image (HSI) classification tasks because of their excellent local spatial feature extraction capabilities. However, it is difficult to establish dependencies between long sequences data for CNNs, there are limitations the process processing spectral sequence features. To overcome these limitations, inspired by Transformer model, a spatial–spectral transformer with cross-attention (CASST) method proposed. Overall,...

Most existing techniques consider hyperspectral anomaly detection (HAD) as background modeling and search problems in the spatial domain. In this article, we model frequency domain treat a frequency-domain analysis problem. We illustrate that spikes amplitude spectrum correspond to background, Gaussian low-pass filter performing on is equivalent an detector. The initial map obtained by reconstruction with filtered raw phase spectrum. To further suppress nonanomaly high-frequency detailed...

With the increasing spectral dimension of hyperspectral images (HSI), how correctly choose bands based on band correlation and information has become more significant, but also complicated. Band selection is a combinatorial optimization problem, intelligent algorithms have been shown to be crucial in solving problems. However, major them only use single objective as index, while neglecting overall features images, which may lead inaccuracy object detection. To tackle this, we propose method...

Hyperspectral images (HSIs) often contain irregular ground cover with mixed spectral features and noise, which makes it challenging to identify the using only pixel features, superpixel or a combination of both. To alleviate above problem, this paper proposes superpixel-pixel-subpixel multilevel network (SPSM), compensates for insufficiencies different levels decrease information loss. For arbitrary regions, are simulated as nodes graph convolutional (GCN) capture spatial texture structure...

In cross-domain hyperspectral image (HSI) classification, the labeled samples of target domain are very limited, and it is a worthy attention to obtain sufficient class information from source categorize classes (both same new unseen classes). This article investigates this problem by employing few-shot learning (FSL) in meta-learning paradigm. However, most existing FSL methods extract statistical features based on convolutional neural networks (CNNs), which typically only consider local...

Deep learning (DL) techniques have shown remarkable progress in remotely sensed hyperspectral image (HSI) classification tasks. The performance of DL-based models highly relies on the quality and quantity labeled data. However, manual labeling is a laborious expensive process that requires substantial efforts from human experts. Active (AL) been developed to alleviate burden annotation by selecting most informative uncertain samples for labeling. In this paper, we propose new class-wise...

Graph theory-based techniques have recently been adopted for anomaly detection in hyperspectral images (HSIs). However, these methods rely excessively on the relational structure within constructed graphs and tend to downplay importance of spectral features original HSI. To address this issue, we introduce graph frequency analysis (HAD), which can serve as a natural tool integrating features. We treat problem location, achieved by constructing beta distribution-based wavelet space, where...

Mislabeled training samples may have a negative effect on the performance of hyperspectral image classification. In order to solve this problem, new density peak (DP) clustering-based noisy label detection method is proposed, which consists following steps. First, distances among each class are calculated using four representative distance metrics, i.e., Euclidean (ED), orthogonal projection divergence (OPD), spectral information (SID), and correlation coefficient (CC). Then, local sample...

Hyperspectral images (HSIs) contain abundant information in the spatial and spectral domains, allowing for a precise characterization of categories materials. Convolutional neural networks (CNNs) have achieved great success HSI classification, owing to their excellent ability local contextual modeling. However, CNNs suffer from fixed filter weights deep convolutional layers, which lead limited receptive field high computational burden. The recent Vision Transformer (ViT) models long-range...

Anomaly detection is a fundamental task in hyperspectral image (HSI) processing. However, most existing methods rely on pixel feature vectors and overlook the relational structure information between pixels, limiting performance. In this article, we propose novel approach to anomaly that characterizes HSI data using vertex-and edge-weighted graph with pixels as vertices. The constructed encodes rich structural an affinity matrix. A crucial innovation of our method ability obtain internal...

Deep neural networks play a significant role in hyperspectral image (HSI) processing, yet they can be easily fooled when trained with adversarial samples (generated by adding tiny perturbations to clean samples). These are invisible the human eye, but lead misclassification deep learning model. Recent research on defense against HSI classification has improved robustness of exploiting global contextual information. However, available methods do not distinguish between different classes...

Abstract Recently, ground coverings change detection (CD) driven by bitemporal hyperspectral images (HSIs) has become a hot topic in the remote sensing community. There are two challenges HSI‐CD task: (1) attribute feature representation of pixel pairs and (2) extraction patterns pairs. To solve above problems, novel spectral‐spatial sequence characteristics‐based convolutional transformer (S3C‐CT) method is proposed for task. In designed method, firstly, an eigenvalue extrema‐based band...