IEEE Geoscience and Remote Sensing Letters

Self-supervised learning (SSL) has shown strong potential for learning representations from unlabeled data, yet its direct application to remote sensing (RS) imagery remains challenging due to complex backgrounds and severe target-scale variations. To address these challenges, we propose Context-Enhanced Semantic Contrast (CESC), a self-supervised pre-training framework tailored for RS imagery. C…

Recent advances in remote sensing image super-resolution (RSISR) have leveraged transformer architectures to achieve state-of-the-art results. Nevertheless, most existing transformer-based approaches treat the token sequence as a unified global context, which often fail to preserve critical local structures and cross-region semantics. To address these issues, we propose an iterative transformer f…

Graph total variation (GTV) has been widely adopted for hyperspectral image (HSI) denoising. However, the fixed graph weights make it difficult to preserve or enhance important details. Moreover, such methods often focus on preserving spatial textures while neglecting the extraction of spectral texture information, failing to achieve a desirable balance between noise removal and spatial-spectral …

Satellite imagery is indispensable for military reconnaissance and environmental monitoring, yet haze pollution severely degrades image quality and target recognition accuracy. Existing methods relying solely on the RGB color space often lead to suboptimal dehazing results with residual haze artifacts. To overcome this limitation, we propose a YCbCr and Depth-guided Collaborative Network (YDCNet)…

Recently, block-term decomposition (BTD) has been successfully applied to hyperspectral image (HSI) processing. In particular, the <italic xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">rank-(L, L, 1)</i> BTD, also known as the LL1 decomposition, decomposes the HSI tensor as a sum of component tensors formed by matrix–vector outer products of <italic xml…

Since multichannel sliding spotlight synthetic aperture radar (SAR) can simultaneously achieve high-resolution and wide-swath (HRWS) remote sensing observations, the large volume of data collected from multiple channels increases the burden on the data link, while the estimation and compensation of channel errors further complicate the signal processing. To reduce hardware costs and algorithm com…

Accurate oriented object detection in aerial imagery is crucial for remote sensing applications, yet remains challenging due to arbitrary orientations, extreme aspect ratios, and dense distributions of objects. Existing rotated detectors often suffer from unstable training gradients and inadequate geometric representation in Intersection over Union (IoU) computation. This study proposes PWIoU-YOL…

Airborne Global Navigation Satellite System-Reflectometry (GNSS-R) offers high-resolution observations but is constrained by time-varying satellite geometries, and currently lacks dedicated trajectory planning methods. To address this, this letter proposes a multi-objective trajectory planning method to balance the trade-off between coverage rate (<italic xmlns:mml="http://www.w3.org/1998/Math/Ma…

Unsupervised band selection aims to identify the most informative spectral bands from hyperspectral images, thereby reducing spectral redundancy while preserving essential spatial–spectral structures. However, existing methods often ignore global or local structural information and noise, leading to suboptimal band subsets. In this paper, we propose a bilateral reweighted reconstruction represent…

Graph-based techniques have shown significant potential in the field of unsupervised change detection for synthetic aperture radar (SAR) imagery. Nonetheless, most existing graph-based approaches rely on a single type of basic unit and perform direct comparisons of weight matrices, which limits their ability to represent multi-scale structural variations. In this paper, we introduce an unsupervis…

Seismic data acquisition is often affected by complex field conditions and equipment precision limitations, leading to the generation of multiple types of noise that reduce the accuracy of reservoir inversion and geological interpretation. Deep learning techniques have begun to be applied in seismic data denoising due to their efficient feature learning capabilities. However, the complex characte…

Synthetic aperture radar (SAR) imagery plays a pivotal role in remote sensing due to its all-weather and all-time imaging capabilities. Despite its advantages, deploying SAR object detection models on resource-limited platforms remains challenging. Existing knowledge distillation methods, primarily designed for natural images, fail to address the unique characteristics of SAR imagery, such as pro…

The two-dimensional L-band Aperture Synthesis Microwave Radiometer (LASMR) aboard China's HY-4 ocean salinity satellite faces significant on-orbit calibration challenges, primarily from ionospheric disturbances, solar radiation, and radio frequency interference (RFI). To address this, we develop a novel screening framework that leverages the instrument's full-polarization measurement capability. …

Spaceborne passive microwave data are known to underestimate both snow cover extent (SCE) and snow water equivalent (SWE). Sensor viewing geometry and scan azimuth angle, in particular, have not been investigated thus far regarding their impact on mapping snow on the ground. In this study, we apply a dry snow detection algorithm to WindSat Level 1C brightness temperatures of 18.7 and 37.0 GHz for…

Our study proposes a new approach for building extraction by introducing the Deformable-Convolution Fusion Feature module, based on the Dual-Attention Network and deformable convolution. The core of the algorithm consists of three modules: Auxiliary Encoder, Main Encoder, and Decoder. The Auxiliary Encoder is responsible for linearly embedding and downsampling the input RGB images and utilizes mu…

Soil moisture is a crucial variable in the climate system. This letter undertakes the retrieval of soil moisture using data from SMOS, which is equipped with an L-band microwave synthetic aperture radiometer. The multi-angle brightness temperature (TB) of the synthetic aperture radiometer is affected, to varying degrees, by factors such as the aliasing effects, ionospheric disturbance, and radio …

Balancing long-range dependencies and efficiency is crucial for remote sensing segmentation. While State Space Models (SSMs) offer linear complexity, they often struggle with local details. To address this, we propose the Lightweight Attention-Mamba Network (LANet). It incorporates an Attention-Guided State Space Module (AG-SSM) to transform indiscriminate scanning into selective refinement of sa…

Hyperspectral anomaly detection (HAD) aims to identify anomalous targets deviating from the background in unlabeled hyperspectral images. Self-supervised methods typically model the background using selected training samples and detect anomalies from reconstruction residuals, but they often weaken target signals, resulting in low detection accuracy. Supervised approaches using annotated or simula…

Wind direction inversion in high-frequency surface wave radar (HFSWR) relies heavily on the directional distribution model (DDM). The classical Longuet-Higgins (L-H) model causes non-physical spatial discontinuities due to energy truncation in the upwind direction. While the Donelan model corrects this defect and achieves higher accuracy, its inversion still relies on non-linear optimization or n…

Unsupervised domain adaptation for object detection in remote sensing images is currently a research hotspot. The significant distribution differences between the source and target domain are the core problem of unsupervised domain adaptation for object detection. Remote sensing images are characterized by large variations in scale, complex backgrounds, and numerous small objects, which pose grea…