SpectralGPT: Spectral Remote Sensing Foundation Model.

The foundation model has recently garnered significant attention due to its potential to revolutionize the field of visual representation learning in a self-supervised manner. While most foundation models are tailored to effectively process RGB images for various visual tasks, there is a noticeable gap in research focused on spectral data, which offers valuable information for scene understanding, especially in remote sensing (RS) applications. To fill this gap, we created for the first time a universal RS foundation model, named SpectralGPT, which is purpose-built to handle spectral RS images using a novel 3D generative pretrained transformer (GPT). Compared to existing foundation models, SpectralGPT 1) accommodates input images with varying sizes, resolutions, time series, and regions in a progressive training fashion, enabling full utilization of extensive RS Big Data; 2) leverages 3D token generation for spatial-spectral coupling; 3) captures spectrally sequential patterns via multi-target reconstruction; 4) trains on one million spectral RS images, yielding models with over 600 million parameters. Our evaluation highlights significant performance improvements with pretrained SpectralGPT models, signifying substantial potential in advancing spectral RS Big Data applications within the field of geoscience across four downstream tasks: single/multi-label scene classification, semantic segmentation, and change detection.

A General Spatial-Frequency Learning Framework for Multimodal Image Fusion.

multimodal image fusion involves tasks like pan-sharpening and depth super-resolution. Both tasks aim to generate high-resolution target images by fusing the complementary information from the texture-rich guidance and low-resolution target counterparts. They are inborn with reconstructing high-frequency information. Despite their inherent frequency domain connection, most existing methods only operate solely in the spatial domain and rarely explore the solutions in the frequency domain. This study addresses this limitation by proposing solutions in both the spatial and frequency domains. We devise a Spatial-Frequency Information Integration Network, abbreviated as SFINet for this purpose. The SFINet includes a core module tailored for image fusion. This module consists of three key components: a spatial-domain information branch, a frequency-domain information branch, and a dual-domain interaction. The spatial-domain information branch employs the spatial convolution-equipped invertible neural operators to integrate local information from different modalities in the spatial domain. Meanwhile, the frequency-domain information branch adopts a modality-aware deep Fourier transformation to capture the image-wide receptive field for exploring global contextual information. In addition, the dual-domain interaction facilitates information flow and the learning of complementary representations. We further present an improved version of SFINet, SFINet++, that enhances the representation of spatial information by replacing the basic convolution unit in the original spatial domain branch with the information-lossless invertible neural operator. We conduct extensive experiments to validate the effectiveness of the proposed networks and demonstrate their outstanding performance against state-of-the-art methods in two representative multimodal image fusion tasks: pan-sharpening and depth super-resolution. The source code is publicly available at https://github.com/manman1995/Awaresome-pansharpening.

Weakly Supervised Solar Panel Mapping via Uncertainty Adjusted Label Transition in Aerial Images.

This paper proposes a novel uncertainty-adjusted label transition (UALT) method for weakly supervised solar panel mapping (WS-SPM) in aerial Images. In weakly supervised learning (WSL), the noisy nature of pseudo labels (PLs) often leads to poor model performance. To address this problem, we formulate the task as a label-noise learning problem and build a statistically consistent mapping model by estimating the instance-dependent transition matrix (IDTM). We propose to estimate the IDTM with a parameterized label transition network describing the relationship between the latent clean labels and noisy PLs. A trace regularizer is employed to impose constraints on the form of IDTM for its stability. To further reduce the estimation difficulty of IDTM, we incorporate uncertainty estimation to first improve the accuracy of noisy dataset distillation and then mitigate the negative impacts of falsely distilled examples with an uncertainty-adjusted re-weighting strategy. Extensive experiments and ablation studies on two challenging aerial data sets support the validity of the proposed UALT.

Bayesian Nonlocal Patch Tensor Factorization for Hyperspectral Image Super-Resolution.

The synthesis of high-resolution (HR) hyperspectral image (HSI) by fusing a low-resolution HSI with a corresponding HR multispectral image has emerged as a prevalent HSI super-resolution (HSR) scheme. Recent researches have revealed that tensor analysis is an emerging tool for HSR. However, most off-the-shelf tensor-based HSR algorithms tend to encounter challenges in rank determination and modeling capacity. To address these issues, we construct nonlocal patch tensors (NPTs) and characterize low-rank structures with coupled Bayesian tensor factorization. It is worth emphasizing that the intrinsic global spectral correlation and nonlocal spatial similarity can be simultaneously explored under the proposed model. Moreover, benefiting from the technique of automatic relevance determination, we propose a hierarchical probabilistic framework based on Canonical Polyadic (CP) factorization, which incorporates a sparsity-inducing prior over the underlying factor matrices. We further develop an effective expectation-maximization-type optimization scheme for framework estimation. In contrast to existing works, the proposed model can infer the latent CP rank of NPT adaptively without tuning parameters. Extensive experiments on synthesized and real datasets illustrate the intrinsic capability of our model in rank determination as well as its superiority in fusion performance.

Feature Pyramid Fusion Network for Hyperspectral Pansharpening.

Hyperspectral (HS) pansharpening aims at fusing an observed HS image with a panchromatic (PAN) image, to produce an image with the high spectral resolution of the former and the high spatial resolution of the latter. Most of the existing convolutional neural networks (CNNs)-based pansharpening methods reconstruct the desired high-resolution image from the encoded low-resolution (LR) representation. However, the encoded LR representation captures semantic information of the image and is inadequate in reconstructing fine details. How to effectively extract high-resolution and LR representations for high-resolution image reconstruction is the main objective of this article. In this article, we propose a feature pyramid fusion network (FPFNet) for pansharpening, which permits the network to extract multiresolution representations from PAN and HS images in two branches. The PAN branch starts from the high-resolution stream that maintains the spatial resolution of the PAN image and gradually adds LR streams in parallel. The structure of the HS branch remains highly consistent with that of the PAN branch, but starts with the LR stream and gradually adds high-resolution streams. The representations with corresponding resolutions of PAN and HS branches are fused and gradually upsampled in a coarse to fine manner to reconstruct the high-resolution HS image. Experimental results on three datasets demonstrate the significant superiority of the proposed FPFNet over the state-of-the-art methods in terms of both qualitative and quantitative comparisons.

A 10-m resolution impervious surface area map for the greater Mekong subregion from remote sensing images.

High-resolution and multi-temporal impervious surface area maps are crucial for capturing rapidly developing urbanization patterns. However, the currently available relevant maps for the greater Mekong subregion suffer from coarse resolution and low accuracy. Addressing this issue, our study focuses on the development of accurate impervious surface area maps at 10-m resolution for this region for the period 2016-2022. To accomplish this, we present a new machine-learning framework implemented on the Google Earth Engine platform that merges Sentinel-1 Synthetic Aperture Radar images and Sentinel-2 Multispectral images to extract impervious surfaces. Furthermore, we also introduce a training sample migration strategy that eliminates the collection of additional training samples and automates multi-temporal impervious surface area mapping. Finally, we perform a quantitative assessment with validation samples interpreted from Google Earth. Results show that the overall accuracy and kappa coefficient of the final impervious surface area maps range from 92.75% to 92.93% and 0.854 to 0.857, respectively. This dataset provides comprehensive measurements of impervious surface coverage and configuration that will help to inform urban studies.

Hyperspectral Image Classification With Multi-Attention Transformer and Adaptive Superpixel Segmentation-Based Active Learning.

Deep learning (DL) based methods represented by convolutional neural networks (CNNs) are widely used in hyperspectral image classification (HSIC). Some of these methods have strong ability to extract local information, but the extraction of long-range features is slightly inefficient, while others are just the opposite. For example, limited by the receptive fields, CNN is difficult to capture the contextual spectral-spatial features from a long-range spectral-spatial relationship. Besides, the success of DL-based methods is greatly attributed to numerous labeled samples, whose acquisition are time-consuming and cost-consuming. To resolve these problems, a hyperspectral classification framework based on multi-attention Transformer (MAT) and adaptive superpixel segmentation-based active learning (MAT-ASSAL) is proposed, which successfully achieves excellent classification performance, especially under the condition of small-size samples. Firstly, a multi-attention Transformer network is built for HSIC. Specifically, the self-attention module of Transformer is applied to model long-range contextual dependency between spectral-spatial embedding. Moreover, in order to capture local features, an outlook-attention module which can efficiently encode fine-level features and contexts into tokens is utilized to improve the correlation between the center spectral-spatial embedding and its surroundings. Secondly, aiming to train a excellent MAT model through limited labeled samples, a novel active learning (AL) based on superpixel segmentation is proposed to select important samples for MAT. Finally, to better integrate local spatial similarity into active learning, an adaptive superpixel (SP) segmentation algorithm, which can save SPs in uninformative regions and preserve edge details in complex regions, is employed to generate better local spatial constraints for AL. Quantitative and qualitative results indicate that the MAT-ASSAL outperforms seven state-of-the-art methods on three HSI datasets.

Noise Prior Knowledge Informed Bayesian Inference Network for Hyperspectral Super-Resolution.

Well-known deep learning (DL) is widely used in fusion based hyperspectral image super-resolution (HS-SR). However, DL-based HS-SR models have been designed mostly using off-the-shelf components from current deep learning toolkits, which lead to two inherent challenges: i) they have largely ignored the prior information contained in the observed images, which may cause the output of the network to deviate from the general prior configuration; ii) they are not specifically designed for HS-SR, making it hard to intuitively understand its implementation mechanism and therefore uninterpretable. In this paper, we propose a noise prior knowledge informed Bayesian inference network for HS-SR. Instead of designing a "black-box" deep model, our proposed network, termed as BayeSR, reasonably embeds the Bayesian inference with the Gaussian noise prior assumption to the deep neural network. In particular, we first construct a Bayesian inference model with the Gaussian noise prior assumption that can be solved iteratively by the proximal gradient algorithm, and then convert each operator involved in the iterative algorithm into a specific form of network connection to construct an unfolding network. In the process of network unfolding, based on the characteristics of the noise matrix, we ingeniously convert the diagonal noise matrix operation which represents the noise variance of each band into the channel attention. As a result, the proposed BayeSR explicitly encodes the prior knowledge possessed by the observed images and considers the intrinsic generation mechanism of HS-SR through the whole network flow. Qualitative and quantitative experimental results demonstrate the superiority of the proposed BayeSR against some state-of-the-art methods.

Dual-Stage Approach Toward Hyperspectral Image Super-Resolution.

Hyperspectral image produces high spectral resolution at the sacrifice of spatial resolution. Without reducing the spectral resolution, improving the resolution in the spatial domain is a very challenging problem. Motivated by the discovery that hyperspectral image exhibits high similarity between adjacent bands in a large spectral range, in this paper, we explore a new structure for hyperspectral image super-resolution (DualSR), leading to a dual-stage design, i.e., coarse stage and fine stage. In coarse stage, five bands with high similarity in a certain spectral range are divided into three groups, and the current band is guided to study the potential knowledge. Under the action of alternative spectral fusion mechanism, the coarse SR image is super-resolved in band-by-band. In order to build model from a global perspective, an enhanced back-projection method via spectral angle constraint is developed in fine stage to learn the content of spatial-spectral consistency, dramatically improving the performance gain. Extensive experiments demonstrate the effectiveness of the proposed coarse stage and fine stage. Besides, our network produces state-of-the-art results against existing works in terms of spatial reconstruction and spectral fidelity. Our code is publicly available at https://github.com/qianngli/DualSR.

Semi-Supervised Multiscale Dynamic Graph Convolution Network for Hyperspectral Image Classification.

In recent years, convolutional neural networks (CNNs)-based methods achieve cracking performance on hyperspectral image (HSI) classification tasks, due to its hierarchical structure and strong nonlinear fitting capacity. Most of them, however, are supervised approaches that need a large number of labeled data to train them. Conventional convolution kernels are fixed shape of rectangular with fixed sizes, which are good at capturing short-range relations between pixels within HSIs but ignore the long-range context within HSIs, limiting their performance. To overcome the limitations mentioned above, we present a dynamic multiscale graph convolutional network (GCN) classifier (DMSGer). DMSGer first constructs a relatively small graph at region-level based on a superpixel segmentation algorithm and metric-learning. A dynamic pixel-level feature update strategy is then applied to the region-level adjacency matrix, which can help DMSGer learn the pixel representation dynamically. Finally, to deeply understand the complex contents within HSIs, our model is expanded into a multiscale version. On the one hand, by introducing graph learning theory, DMSGer accomplishes HSI classification tasks in a semi-supervised manner, relieving the pressure of collecting abundant labeled samples. Superpixels are generally in irregular shapes and sizes which can group only similar pixels in a neighborhood. On the other hand, based on the proposed dynamic-GCN, the pixel-level and region-level information can be captured simultaneously in one graph convolution layer such that the classification results can be improved. Also, due to the proper multiscale expansion, more helpful information can be captured from HSIs. Extensive experiments were conducted on four public HSIs, and the promising results illustrate that our DMSGer is robust in classifying HSIs. Our source codes are available at https://github.com/TangXu-Group/DMSGer.

Graph-in-Graph Convolutional Network for Hyperspectral Image Classification.

With the development of hyperspectral sensors, accessible hyperspectral images (HSIs) are increasing, and pixel-oriented classification has attracted much attention. Recently, graph convolutional networks (GCNs) have been proposed to process graph-structured data in non-Euclidean domains and have been employed in HSI classification. But most methods based on GCN are hard to sufficiently exploit information of ground objects due to feature aggregation. To solve this issue, in this article, we proposed a graph-in-graph (GiG) model and a related GiG convolutional network (GiGCN) for HSI classification from a superpixel viewpoint. The GiG representation covers information inside and outside superpixels, respectively, corresponding to the local and global characteristics of ground objects. Concretely, after segmenting HSI into disjoint superpixels, each one is converted to an internal graph. Meanwhile, an external graph is constructed according to the spatial adjacent relationships among superpixels. Significantly, each node in the external graph embeds a corresponding internal graph, forming the so-called GiG structure. Then, GiGCN composed of internal and External graph convolution (EGC) is designed to extract hierarchical features and integrate them into multiple scales, improving the discriminability of GiGCN. Ensemble learning is incorporated to further boost the robustness of GiGCN. It is worth noting that we are the first to propose the GiG framework from the superpixel point and the GiGCN scheme for HSI classification. Experiment results on four benchmark datasets demonstrate that our proposed method is effective and feasible for HSI classification with limited labeled samples. For study replication, the code developed for this study is available at https://github.com/ShuGuoJ/GiGCN.git.

SpaSSA: Superpixelwise Adaptive SSA for Unsupervised Spatial-Spectral Feature Extraction in Hyperspectral Image.

Singular spectral analysis (SSA) has recently been successfully applied to feature extraction in hyperspectral image (HSI), including conventional (1-D) SSA in spectral domain and 2-D SSA in spatial domain. However, there are some drawbacks, such as sensitivity to the window size, high computational complexity under a large window, and failing to extract joint spectral-spatial features. To tackle these issues, in this article, we propose superpixelwise adaptive SSA (SpaSSA), that is superpixelwise adaptive SSA for exploiting local spatial information of HSI. The extraction of local (instead of global) features, particularly in HSI, can be more effective for characterizing the objects within an image. In SpaSSA, conventional SSA and 2-D SSA are combined and adaptively applied to each superpixel derived from an oversegmented HSI. According to the size of the derived superpixels, either SSA or 2-D singular spectrum analysis (2D-SSA) is adaptively applied for feature extraction, where the embedding window in 2D-SSA is also adaptive to the size of the superpixel. Experimental results on the three datasets have shown that the proposed SpaSSA outperforms both SSA and 2D-SSA in terms of classification accuracy and computational complexity. By combining SpaSSA with the principal component analysis (SpaSSA-PCA), the accuracy of land-cover analysis can be further improved, outperforming several state-of-the-art approaches.

Moving Vehicle Detection for Remote Sensing Video Surveillance With Nonstationary Satellite Platform.

With satellite platforms gazing at a target territory, the captured satellite videos exhibit local misalignment and local intensity variation on some stationary objects that can be mistakenly extracted as moving objects and increase false alarm rates. Typical approaches for mitigating the effect of moving cameras in moving object detection (MOD) follow domain transformation technique, where the misalignment between consecutive frames is restricted to the image planar. However, such technique cannot properly handle satellite videos, as the local misalignment on them is caused by the varying projections from the 3D objects on the Earth's surface to 2D image planar. In order to suppress the effect of moving satellite platform in MOD, we propose a Moving-Confidence-Assisted Matrix Decomposition (MCMD) model, where foreground regularization is designed to promote real moving objects and ignore system movements with the assistance of a moving-confidence score estimated from dense optical flows. For solving the convex optimization problem in MCMD, both batch processing and online solutions are developed in this study, by adopting the alternating direction method and the stochastic optimization strategy, respectively. Experimental results on the videos captured by SkySat and Jilin-1 show that MCMD outperforms the state-of-the-art techniques with improved precision by suppressing effect of nonstationary satellite platforms.

Self-spectral learning with GAN based spectral-spatial target detection for hyperspectral image.

To alleviate the shortcomings of target detection in only one aspect and reduce redundant information among adjacent bands, we propose a spectral-spatial target detection (SSTD) framework in deep latent space based on self-spectral learning (SSL) with a spectral generative adversarial network (GAN). The concept of SSL is introduced into hyperspectral feature extraction in an unsupervised fashion with the purpose of background suppression and target saliency. In particular, a novel structure-to-structure selection rule that takes full account of the structure, contrast, and luminance similarity is established to interpret the mapping relationship between the latent spectral feature space and the original spectral band space, to generate the optimal spectral band subset without any prior knowledge. Finally, the comprehensive result is achieved by nonlinearly combining the spatial detection on the fused latent features with the spectral detection on the selected band subset and the corresponding selected target signature. This paper paves a novel self-spectral learning way for hyperspectral target detection and identifies sensitive bands for specific targets in practice. Comparative analyses demonstrate that the proposed SSTD method presents superior detection performance compared with CSCR, ACE, CEM, hCEM, and ECEM.

Deep learning based regression for optically inactive inland water quality parameter estimation using airborne hyperspectral imagery.

Airborne hyperspectral remote sensing has the characteristics of high spatial and spectral resolutions, and provides an opportunity for accurate and efficient inland water qauality monitoring. Many studies have focused on evaluating and quantifying the concentrations of the optically active water quality parameters, for parameters such as chlorophyll-a (Chla), cyanobacteria, and colored dissolved organic matter (CDOM). For the optically inactive parameters, such as the permanganate index (CODMn), total nitrogen (TN), total phosphorus (TP), ammoniacal nitrogen (NH3-N), and heavy metals, it is difficult to estimate the concentrations directly, and the traditional indirect estimation models cannot meet the accuracy requirements, especially in heavily polluted inland waters. In this study, 60 water samples were collected at a depth of 50 cm from the Guanhe River in China, at the same time as the airborne data acquisition. We also developed and investigated two deep learning based regression models-a pixel-based deep neural network regression (pixel_DNNR) model and a patch-based deep neural network regression (patch_DNNR) model-to estimate seven optically inactive water quality parameters. Compared with the partial least squares regression (PLSR) and support vector regression (SVR) models, the deep learning based regression models can obtain a superior accuracy, especially the patch_DNNR model, which obtained a superior prediction accuracy for all parameters, with the prediction dataset coefficient of determination (Rp2) and the residual prediction deviation (RPD) values being greater than 0.6 and 1.6, respectively. In addition, thematic maps of the water quality classification results and water parameter concentrations were generated and the overall water quality and pollution sources were analyzed in the study area. The experimental results demonstrate that the deep learning based regression models show a good performance in the feature extraction and image understanding of high-dimensional data, and they provide us with a new approach for optically inactive inland water quality parameter estimation.

Unsupervised spectral mapping and feature selection for hyperspectral anomaly detection.

Exploring techniques that breakthrough the unknown space or material species is of considerable significance to military and civilian fields, and it is a challenging task without any prior information. Nowadays, the use of material-specific spectral information to detect unknowns has received increasing interest. However, affected by noise and interference, high-dimensional hyperspectral anomaly detection is difficult to meet the requirements of high detection accuracy and low false alarm rate. Besides, there is a problem of insufficient and unbalanced samples. To address these problems, we propose a novel hyperspectral anomaly detection framework based on spectral mapping and feature selection (SMFS) in an unsupervised manner. The SMFS introduces the essential properties of hyperspectral data into an unsupervised neural network to construct the nonlinear mapping relationship from high-dimensional spectral space to low-dimensional deep feature space. And it searches the optimal feature subset from the candidate feature space for standing out anomalies. Because of the compelling characterization of the encoder, we develop it specifically for spectral signatures to reveal the hidden data. Quantitative and qualitative experiments on real hyperspectral datasets indicate that the proposed method can provide the compact features overcoming the problems of noise, interference, redundancy and time-consuming caused by high-dimensionality and limited samples. And it has advantages over some state-of-the-art competitors concerning detecting anomalies of different scales.

Spectral constraint adversarial autoencoders approach to feature representation in hyperspectral anomaly detection.

Anomaly detection in hyperspectral images (HSIs) faces various levels of difficulty due to the high dimensionality, redundant information and deteriorated bands. To address these problems, we propose a novel unsupervised feature representation approach by incorporating a spectral constraint strategy into adversarial autoencoders (AAE) without any prior knowledge in this paper. Our approach, called SC_AAE (spectral constraint AAE), is based on the characteristics of HSIs to obtain better discrimination represented by hidden nodes. To be specific, we adopt a spectral angle distance into the loss function of AAE to enforce spectral consistency. Considering the different contribution rates of each hidden node to anomaly detection, we individually fuse the hidden nodes by an adaptive weighting method. A bi-layer architecture is then designed to suppress the variational background (BKG) while preserving features of anomalies. The experimental results demonstrate that our proposed method outperforms the state-of-the-art methods.

Incorporating Negative Sample Training for Ship Detection Based on Deep Learning.

While ship detection using high-resolution optical satellite images plays an important role in various civilian fields-including maritime traffic survey and maritime rescue-it is a difficult task due to influences of the complex background, especially when ships are near to land. In current literatures, land masking is generally required before ship detection to avoid many false alarms on land. However, sea⁻land segmentation not only has the risk of segmentation errors, but also requires expertise to adjust parameters. In this study, Faster Region-based Convolutional Neural Network (Faster R-CNN) is applied to detect ships without the need for land masking. We propose an effective training strategy for the Faster R-CNN by incorporating a large number of images containing only terrestrial regions as negative samples without any manual marking, which is different from the selection of negative samples by targeted way in other detection methods. The experiments using Gaofen-1 satellite (GF-1), Gaofen-2 satellite (GF-2), and Jilin-1 satellite (JL-1) images as testing datasets under different ship detection conditions were carried out to evaluate the effectiveness of the proposed strategy in the avoidance of false alarms on land. The results show that the method incorporating negative sample training can largely reduce false alarms in terrestrial areas, and is superior in detection performance, algorithm complexity, and time consumption. Compared with the method based on sea⁻land segmentation, the proposed method achieves the absolute increment of 70% of the F1-measure, when the image contains large land area such as the GF-1 image, and achieves the absolute increment of 42.5% for images with complex harbors and many coastal ships, such as the JL-1 images.

Effective Sequential Classifier Training for SVM-Based Multitemporal Remote Sensing Image Classification.

The explosive availability of remote sensing images has challenged supervised classification algorithms such as Support Vector Machines (SVM), as training samples tend to be highly limited due to the expensive and laborious task of ground truthing. The temporal correlation and spectral similarity between multitemporal images have opened up an opportunity to alleviate this problem. In this study, a SVM-based Sequential Classifier Training (SCT-SVM) approach is proposed for multitemporal remote sensing image classification. The approach leverages the classifiers of previous images to reduce the required number of training samples for the classifier training of an incoming image. For each incoming image, a rough classifier is firstly predicted based on the temporal trend of a set of previous classifiers. The predicted classifier is then fine-tuned into a more accurate position with current training samples. This approach can be applied progressively to sequential image data, with only a small number of training samples being required from each image. Experiments were conducted with Sentinel-2A multitemporal data over an agricultural area in Australia. Results showed that the proposed SCT-SVM achieved better classification accuracies compared with two state-of-the-art model transfer algorithms. When training data are insufficient, the overall classification accuracy of the incoming image was improved from 76.18% to 94.02% with the proposed SCT-SVM, compared with those obtained without the assistance from previous images. These results demonstrate that the leverage of a priori information from previous images can provide advantageous assistance for later images in multitemporal image classification.

A Dynamic Neighborhood Learning-Based Gravitational Search Algorithm.

Balancing exploration and exploitation according to evolutionary states is crucial to meta-heuristic search (M-HS) algorithms. Owing to its simplicity in theory and effectiveness in global optimization, gravitational search algorithm (GSA) has attracted increasing attention in recent years. However, the tradeoff between exploration and exploitation in GSA is achieved mainly by adjusting the size of an archive, named , which stores those superior agents after fitness sorting in each iteration. Since the global property of remains unchanged in the whole evolutionary process, GSA emphasizes exploitation over exploration and suffers from rapid loss of diversity and premature convergence. To address these problems, in this paper, we propose a dynamic neighborhood learning (DNL) strategy to replace the model and thereby present a DNL-based GSA (DNLGSA). The method incorporates the local and global neighborhood topologies for enhancing the exploration and obtaining adaptive balance between exploration and exploitation. The local neighborhoods are dynamically formed based on evolutionary states. To delineate the evolutionary states, two convergence criteria named limit value and population diversity, are introduced. Moreover, a mutation operator is designed for escaping from the local optima on the basis of evolutionary states. The proposed algorithm was evaluated on 27 benchmark problems with different characteristic and various difficulties. The results reveal that DNLGSA exhibits competitive performances when compared with a variety of state-of-the-art M-HS algorithms. Moreover, the incorporation of local neighborhood topology reduces the numbers of calculations of gravitational force and thus alleviates the high computational cost of GSA.