ABSTRACT With the rapid advancement of communication technologies, space‐air‐ground integrated networks (SAGIN) have become a pivotal research frontier in current and future communication domains. To tackle critical challenges in SAGIN scenarios, such as excessive task‐related energy consumption and insufficient communication‐computing resources, this paper proposes a three‐tier edge computing architecture integrating satellites, unmanned aerial vehicle (UAV) swarms, and ground systems. Aiming to minimize the system's weighted energy consumption and latency, we investigate the joint optimization of task allocation, user‐UAV association, UAV deployment, and resource allocation between UAVs and low‐earth orbit (LEO) satellites. Formulated as a non‐convex mixed‐integer nonlinear combinatorial optimization problem, this work integrates the branch‐and‐bound method, multi‐start global optimization, and gray wolf optimization (GWO) to develop a suboptimal solution based on block coordinate descent (BCD), which decouples the original problem into three subproblems for independent solving and iterative approximation of the optimal solution. Experimental results show that the proposed algorithm reduces the total system cost by 7.81%, 11.99%, and 45.69% compared with baseline algorithms with random user‐UAV association, random UAV positioning, and random task assignment, respectively, effectively cutting down overall energy consumption and task latency.

On-orbit rapid detection of land surface anomalies is important for ensuring ecological security and human safety. Land surface anomalies (e.g., fire, industrial heat source, and deforestation, etc.) are often accompanied by different degrees of thermal anomalies. Existing methods for detecting thermal anomalies have focused primarily on high-temperature anomalies, without available approach for detecting widespread low-temperature anomalies. Here, a Novel Method based on Constructed Reference land surface temperatures (LST) for on-orbit remote sensing detection of various Thermal Anomalies (NMCRTA) is proposed and further evaluated using Landsat 8 LST product. In this method, we first construct an fitted reference temperature based on LST spatiotemporal trend surface modeling and a real reference temperature based on contextual averaging. Then, the difference (including the step of removing atmospheric effects) between on-orbit observed LST and fitted reference LST, and the difference between on-orbit observed LST and real reference LST are calculated, respectively. Finally, these two temperature differences are utilized to detect thermal anomalies using corresponding thresholds. The results indicate that the NMCRTA can effectively detect deforestation, newly constructed buildings, and river drying, with an overall F1-score of 0.867, in a 100 × 100 km region scale. Meanwhile, the NMCRTA exhibited excellent accuracy in detecting fires, deforestation, and landslides at the 15 × 15 km scene scale, achieving F1-scores of 0.943, 0.857, and 0.791, respectively. Furthermore, the NMCRTA can continuously capture different thermal anomaly events associated with a newly constructed industrial heat source and perform well in nighttime. The NMCRTA is promising for future on-orbit remote sensing detection of various land surface anomalies, as a valuable supplement to optical on-orbit detection method.

Deep learning-based classification methods of hyperspectral images (HSIs) have made significant progress recently, catching the attention of academia and industry; however, the existing studies of classification of HSIs mainly focus on the closed-set environment with the assumption that ground classes are fixed and known, ignoring the complexity and diversity of ground objects in the real world. As a result, the unknown classes will be forced into known classes. To solve this problem, we propose a novel spectral-spatial evidential learning (SSEL) network that combines an improved generative adversarial network (GAN) and evidential theory for open-set classification of HSIs. First, a domain adaptation (DA) strategy is embedded into GAN to generate high-quality samples by reducing the distribution discrepancy between generated and real samples. Second, the discriminator is devised to extract spectral-spatial features and output multiclass evidence for closed-set classification and uncertainty estimation. A new classification function called evidence-based loss is designed for the discriminator to guide the evidence-collection process. Additionally, a novel adversarial objective function is defined, where the discriminator loss is devised to predict real samples belonging to the true class and generated samples belonging to “none of the classes. The generator loss is developed to generate samples consistent with the label category. Finally, the class and corresponding uncertainty can be calculated based on the collected evidence and the appropriate open-set classification of HSIs. Extensive experiments on three benchmark HSIs show that our proposed method achieves competitive performance on closed-set and open-set classification of HSIs compared with existing state-of-the-art methods.

Unseen category prediction is a common challenge for real-world applications, especially for remote sensing (RS) imagery interpretation. Zero-shot learning (ZSL)--based scene classification methods have made significant progress recently, providing an effective solution for unseen scene recognition with semantic embeddings that link seen and unseen classes in the field of RS. However, existing ZSL methods mainly focus on semantic feature exploration, they failed to combine image features and semantic features effectively. To address the aforementioned challenges, we propose a novel knowledge graph-based zero-shot learning model that adeptly integrates both image and semantic features for disaster RS scene recognition. First, we construct an RS knowledge graph to generate semantic features of RS scenes, enhancing the reasoning ability from conventional RS scene categories to disaster RS scene categories. Second, we propose an Interactive Attention mechanism to integrate image and semantic features, focusing on the most informative regions. Finally, we introduce an RS domain adapter that enables the model to better adapt to remote sensing data, reproject common features into the remote sensing domain, and thus solve zero-shot remote sensing scene classification tasks. To demonstrate the effectiveness of our method, we construct a remote sensing disaster scene dataset, which contains 8700 high-quality disaster scenes. Extensive experiments show that our proposed method outperforms current state-of-the-art methods under zero-shot RS image scene classification settings.

Early warning systems and emergency management for disasters, environmental pollution, and illegal development require timely and accurate Earth surface anomaly detection (ESAD). Remote sensing, which uses satellites to observe the Earth’s surface, is an emerging approach to address this need. However, current remote sensing methods for ESAD are limited by their focus on specific anomalies, reliance on high-level satellite data, and the demand for significant computational and storage resources. In this article, we present a novel framework for general and on-orbit ESAD, which combines large vision models and lightweight priors. Our method characterizes images with a large vision model that is highly generalizable, reducing the dependency on high-level data, and compressing the prior base with sampling techniques, facilitating transmission and on-orbit storage. For on-orbit detection, we use a dictionary look-up style method for efficient anomaly prediction, enabling detection on satellites with limited computation resources. We evaluate our framework on typical scenarios and compare it with popular change detection (CD)-based and anomaly detection (AD)-based methods. Our results show that our framework achieves good performance while reducing the prior size by at least 95.56 times. Moreover, our framework can handle unpaired data, providing a chance to detect anomalies in the absence of near-term and paired images. Our framework has the potential to support the development and applications of general, on-orbit ESAD. The code and dataset are available at the following site:https://github.com/YummyWaffle/ESAD.

Accurate and timely monitoring of wildfires is crucial for reducing property damage and casualties. In recent years, advances in satellite technology have enabled the comprehensive, timely, and rapid recording of various abrupt events on the Earth’s surface. However, achieving a balance between temporal and spatial resolution remains a significant challenge for remote sensing, hindering the quick and accurate detection of wildfires. This letter proposes a novel framework for the near real-time monitoring of wildfire coupled with the BRDF model and deep learning technology, which enables near real-time detection of wildfire by assessing the degree to which the observed value of geostationary satellite image deviates from the predicted theoretical observation value. The experimental results show that the proposed method is capable of effectively detecting wildfires in near real-time. Moreover, the encouraging results suggest that the method holds promise for monitoring the spread of wildfire to a certain extent.

Time-series image change detection is one of the most challenging tasks to remote sensing society. Due to complex phenological patterns of cropland, it is difficult to design an efficient strategy for cropland change detection. In this work, an integrated framework is proposed to perform change detection with a limited number of training samples. There are two improvements in this proposed cropland change detection method: 1) the harmonic function is utilized to fill the missing data within a time-series image stack by considering phenological patterns of cropland and 2) the CropGAN was developed to generate realistic samples for training data set enrichment. Compared to the traditional change detection methods, the proposed strategy able to detect different kinds of cropland changes even with few number of samples. Experiments on a Landsat time-series image stack demonstrated that the proposed CropGAN can significantly improve change detection accuracies, given a limited number of labeled samples.

Change detection by comparing two bitemporal images is one of the most challenging tasks in remote sensing. At present, most related studies focus on change area detection while neglecting multiple change type identification. In this letter, an attention gates generative adversarial adaptation network (AG-GAAN) is proposed on multiple change detection. The AG-GAAN has the following contributions: 1) this method can automatically detect multiple changes; 2) it includes attention gates mechanism for spatial constraint and accelerates change area identification with finer contours; and 3) the domain similarity loss is introduced to improve the discriminability of the model so that the model can map out real changes more accurately. To demonstrate the robustness of this approach, we used the Google Earth data sets that include seasonal variations for change detection and understanding. The experimental results demonstrated that the proposed method can accurately detect the multiple change types from bitemporal imagery.

The timely and accurate land cover mapping with remote sensing images played a huge role in ecosystem monitoring. However, due to the spectral variability and spatial complexity of high-resolution remote sensing images, it is often difficult to find an efficient method to achieve accurate land cover classification. To explore useful contextual and hierarchical features in remote sensing images, this paper proposes a U-shaped object graph neural network (U-OGNN), which is mainly composed of self-adaptive graph construction (SAGC), hierarchical graph encoder, and decoder. For self-adaptive graph construction, the similarity measurement is applied to generate contextual-aware graph structure, by feeding deep features extracted from convolution and multi-layer attention operations. Graph encoder and decoder fuse multi-level information over different scales by capturing hierarchical features of adjacent objects. In this way, the proposed method is able to predict land-cover types by considering multi-level contextual information accurately. Experiments on GID land-cover classification datasets, the overall accuracies of the U-OGNN reach 87.81 %.

Deep semantic segmentation networks perform better in cloud detection of satellite imagery than traditional methods due to their ability to extract high-level features over a large receptive field. However, a large receptive field often leads to loss of spatial details and blurring of boundaries. Therefore, it is crucial to understand the role of the receptive field on the segmentation results, which has rarely been investigated for cloud detection tasks. This study, for the first time, explored the relationship between the receptive field size and the performance of a cloud detection network. Six typical networks commonly used for cloud detection and nine modified UNet variants with different depths, dilated convolutions, and skip connections were evaluated based on the Landsat 8 Biome (L8 Biome) dataset. The theoretical receptive field (TRF) and the effective receptive field (ERF) were introduced to measure the receptive field sizes of different networks. The results revealed a negative correlation between the ERF size and cloud segmentation accuracies for different cloud distributions and a relatively weak negative correlation between the TRF size and segmentation accuracies. Furthermore, ERFs were considerably smaller than the corresponding TRFs for most networks, implying that large-scale contextual information was not learned after training. This result indicates the importance of using networks with a small receptive field for cloud detection of Landsat 8 OLI imagery. Moreover, as the boundary accuracies are significantly lower than the region accuracies, future efforts should be devoted to addressing inaccurate boundary localization rather than exploring the contextual information over a large receptive field.

The rapid development of hyperspectral remote sensing technology, has led to an explosion in the number of available hyperspectral images (HSI). The fast and accurate characterization of HSI poses a significant challenge for remote sensing scientists. Currently, deep learning strategies with various neural networks have been successfully applied for HSI classification using the concept of the “dataset-model”. Still, there is a need to develop universal deep learning models for HSI classification using a continual updating strategy. This paper presents a life-long learning strategy to continually update model weights with the help of continual spectral-spatial feature distillation. Specifically, the proposed method introduces a spectral-spatial distillation strategy to retain knowledge of the previous well-trained model. Meanwhile, the learning metric term is integrated into a multi-level feature extraction to minimize the spectral-spatial feature discrepancy between the previous model and the new one. The experimental results indicate that our method achieves superior performance for continual HSI classification tasks without suffering from the persistent loss of characterization memory.

Hyperspectral image classification is a challenging task when a limited number of training samples are available. It is also known that the classification performance highly depends on the quality of the labeled samples. In this work, a cluster-based conditional generative adversarial net (CCGAN) is proposed as an effective solution to increase the size and quality of the training data set. The proposed method is able to automatically select the most representative initial samples with a subtractive clustering-based strategy, which keeps the diversity for sample generation. Moreover, compared to the traditional semisupervised classification frameworks, the CCGAN is able to generate realistic spectral profiles by considering the class-specific labels. Experiments on well-known Pavia University data set demonstrate that the proposed CCGAN can significantly boost the classification accuracy, even using a small number of initial labeled samples.

Change detection by comparing two bitemporal images is one of the most fundamental challenges for dynamic monitoring of the Earth surface. In this article, we propose a metric learning-based generative adversarial network (GAN) (MeGAN) to automatically explore seasonal invariant features for pseudochange suppressing and real change detection. To achieve this purpose, a seasonal invariant term is introduced to maximally suppress pseudochanges, whereas the MeGAN explores the transition patterns between adjacent images in a self-learning fashion. Different from the previous works on bitemporal imagery change detection, the proposed MeGAN have the following contributions: 1) it automatically explores change patterns from the complex bitemporal background without human intervention and 2) it aims to maximally exclude pseudochanges from the seasonal transition term and map out real changes efficiently. To our best knowledge, this is the first time we incorporate the seasonal transition term and GAN for change detection between bitemporal images. At last, to demonstrate the robustness of the proposed method, we included two data sets which are the Google Earth data and the Landsat data, for bitemporal change detection and evaluation. The experimental results indicated that the proposed method is able to perform change detection with precision can be as high as 81% and 88% for the Google Earth and Landsat data set, respectively.

In this paper, we propose a spectral–spatial feature based classification (SSFC) framework that jointly uses dimension reduction and deep learning techniques for spectral and spatial feature extraction, respectively. In this framework, a balanced local discriminant embedding algorithm is proposed for spectral feature extraction from high-dimensional hyperspectral data sets. In the meantime, convolutional neural network is utilized to automatically find spatial-related features at high levels. Then, the fusion feature is extracted by stacking spectral and spatial features together. Finally, the multiple-feature-based classifier is trained for image classification. Experimental results on well-known hyperspectral data sets show that the proposed SSFC method outperforms other commonly used methods for hyperspectral image classification.