Search Results (125)

Deep-learning-based land cover segmentation models for high-resolution aerial images often suffer from inconsistent performance on unseen data due to limited spatial and temporal dataset ranges alongside inherent labeling inconsistencies. Existing solutions, such as foundation models and unsupervised domain adaptation, offer indirect improvements but require considerable post-processing or transfer learning. To address this limitation, this study introduces a direct dataset integration method that fuses heterogeneous datasets with disparate class compositions into a single operational framework. For each individual dataset, multiple model instances (MMIs) predict pixel-wise class probabilities. These probabilities are subsequently combined within a single, unified, and customizable label space, enabling improved generalization to unseen domains without additional manual annotation. We applied and validated this method using three benchmark remote sensing datasets: OpenEarthMap, FLAIR, and the cloud dataset based on high-resolution optical satellite imagery (KOMPSAT-3 and 3A), sourced from the official AI-Hub platform in South Korea. Across both 2-fold and holdout tests, statistical validation confirmed that the unified model enhances generalization on untrained imagery (p < 0.0001) while reliably maintaining or slightly improving original intra-dataset accuracy. Furthermore, it successfully consolidates diverse land-cover categories with atmospheric features such as clouds, which were previously handled only by separate dedicated models. Full article

This study benchmarks four classifiers—Random Forest (RF), Classification and Regression Trees (CART), Support Vector Machine (SVM), and Gradient Tree Boosting (GTB)—on an identical 16-feature KOMPSAT-3/5 and Sentinel-1/2 stack harmonized to a common 2.8 m grid within Google Earth Engine. RF and GTB reached overall accuracies of 92.4% and 92.8%, separating them from the unconstrained CART under family-wise correction and from the linear-kernel SVM under false-discovery-rate control. Class-specific SHAP was cross-interpreted against permutation importance, feature-correlation analysis, and a leave-one-sensor-out (LOSO) ablation, with SHAP and permutation-importance rankings in broad agreement (Spearman ρ = 0.72–0.92). The central contribution is to characterize two conditions—strong within-sensor collinearity and class-conditional information concentration—under which feature-level attribution and sensor-level necessity diverge. In the Bare Land class, no individual Sentinel-2 SWIR band ranks among the top three by SHAP, yet withholding the Sentinel-2 group produces the largest class-wise degradation observed (ΔF1 = 0.102 for RF and 0.130 for GTB), consistent with attribution dilution across the collinear SWIR1–SWIR2 pair (r = 0.97). Because none of the sixteen ablation contrasts survives multiplicity correction with a sample size of 250 (n = 250), these contrasts are reported as effect-size estimates rather than confirmatory tests. Full article

High-resolution (HR) Ground Control Point (GCP) chips are essential for accurate satellite image registration, yet their generation commonly relies on aerial imagery, which is often limited by data availability and update frequency. To improve the usability of satellite-derived GCP chips as an alternative, this study systematically investigates the influence of deep learning-based image super-resolution (SR) on GCP chip template matching under varying input spatial resolution conditions. GCP chips extracted from KOMPSAT-3 imagery were synthetically degraded using ×4 and ×2 downsampling, while original-resolution chips (×1) were also included for comparison. The degraded chips were enhanced using SR models and matched to KOMPSAT-3A imagery using both conventional and deep learning-based template matching methods. The experimental results show that SR is more effective for input GCP chips with limited spatial detail, with the largest improvements obtained under the ×4 downsampling condition (~2.8 m GSD). The contribution of SR is further influenced by geometric resampling during template matching and by the adopted matching method. These findings identify the conditions under which SR provides meaningful improvements and offer practical guidance for enhancing satellite-derived GCP chips, extending their usability for satellite image registration when aerial imagery is unavailable or requires frequent updates. Full article

With the rapid growth of the New Space era, the increasing number and diversity of Low Earth Orbit (LEO) satellites demand generic and reusable antenna control software capable of stable and accurate tracking across heterogeneous orbital and antenna environments. Previous works have primarily focused on two-axis antenna systems, relying on step-based tracking or Two-Line Element (TLE)-based orbit prediction. However, such approaches suffer from manual initialization requirements and gimbal lock at high elevation angles, limiting continuous tracking performance. Although three-axis antenna structures have been proposed to address these issues, their lack of software-level generality and experimental validation restricts practical reuse across antenna platforms. This paper presents a generic three-axis parabola antenna control software that integrates TLE-based orbit prediction, a gimbal lock avoidance algorithm, and an antenna specification reflection function. By decoupling antenna-specific parameters from the core tracking logic, the proposed system enables flexible adaptation to diverse antenna configurations. Experimental validation using a real three-axis antenna and simulations demonstrates high tracking accuracy, achieving an average gain error within 0.466 dBm for GEO-KOMPSAT-2A and stable signal strength above 55 dBm for the LEO satellite ARIRANG-5. Furthermore, analysis of 11,469 gimbal-lock-prone satellites confirms robust avoidance performance across diverse orbital conditions. Full article

Spaceborne synthetic aperture radar (SAR) needs point-like reference targets for co-registration, calibration, and absolute geolocation, but real corner reflectors (CRs) exist only at a few dedicated sites. We present a detection and three-dimensional (3D) positioning pipeline for naturally occurring corner-reflector-like (pseudo-CR) scatterers in KOMPSAT-5 stacks, combining constant false alarm rate (CFAR) detection ranked by local contrast rather than absolute power, a multi-scale template bank, multi-look range-Doppler bundle triangulation, and a per-target conditioning diagnostic vector. At the nine-CR Mongolia calibration field, all nine are recovered, with a mean horizontal error of 1.37 m and a mean 3D error of 2.01 m. At a CR-absent suburban site in Suncheon, Republic of Korea, 18 scenes yield 39,822 clusters, 26,659 of which pass the conditioning gate. Over 16 response-enriched poles retained by author adjudication, the difference to the nearest conditioning-filtered cluster averages 3.19 m horizontally and 1.45 m vertically relative to the surveyed pole base. Because part of that set was selected after inspecting preliminary responses, these are conditional nearest-cluster agreements, not verified target accuracies, and no detection probability is claimed. In an exploratory comparison on the same set, contrast ranking retains more evaluation targets than power ranking at equal budgets, 16 against 9 of 16. Removing robust weighting and pruning degrades the mean Mongolia 3D error to 7.39 m; the DEM prior is non-critical, whereas the clustering radius is consequential in the dense pole field, with the count within 20 m falling from 16 to 9 as the radius grows from 15 m to 30 m. In a leave-scene-out check, 85.8% of 254 evaluable clusters persist in a scene excluded from their solutions, against 35.2% of 270 evaluable random positions: a complete-case enrichment of 78%, or 53–82% under cross-arm worst-case imputation of the unevaluable targets; neither estimate measures catalogue precision. Limited KOMPSAT-5 stacks can therefore yield few-metre nearest-cluster agreement at selected pole locations where stable multi-scene responses form. Full article

Multi-sensor fusion of optical and synthetic aperture radar (SAR) imagery is widely used for land cover classification, yet most studies treat heterogeneous sensors as a uniform feature pool, leaving class-dependent differences in sensor contribution insufficiently quantified. To address this gap, this study integrates optical and SAR imagery within a Random Forest (RF) classifier in Google Earth Engine (GEE) and applies a combined SHapley Additive exPlanations (SHAP)—Remove and Retrain (ROAR)—bootstrap framework to disentangle, for each class, which sensor provides which discriminative information and how faithful those attributions are. The dataset included Sentinel-1/2, Korea Multi-Purpose Satellite (KOMPSAT)-3A/5, and Landsat-8, representing a range of spatial resolutions and spectral characteristics. An RF-based machine learning (ML) model was employed to perform multi-sensor data fusion and classification. To address the inherent opacity of ML models, we employed the SHAP algorithm, an explainable artificial intelligence (XAI) method, to interpret classification decisions. SHAP analysis indicated that visible and near-infrared (NIR) bands, along with vegetation indices, were the dominant contributors to land cover classification in this study area, while SAR data provided complementary structural information for spectrally ambiguous targets, such as roads—a class-dependent role reflected in the stable rank ordering of the Sentinel-1 VV contribution across bootstrap replications, rather than in a formally significant magnitude difference. ROAR results were consistent with the top-ranked SHAP features being those on which the classifier relies, supporting a physically plausible interpretation of ML-based remote sensing classification. These attributions were robust to estimator choice (SHAP-permutation ρ = 0.89) and spatial partitioning (mean ρ = 0.97 across folds). The contribution of this study is methodological rather than algorithmic: it provides an integrated analytical framework that applies SHAP, ROAR, and bootstrap confidence intervals jointly to a specific multi-sensor land cover problem, demonstrating that interpretability and high classification accuracy can be reported together. The results demonstrate that optical and SAR fusion contribute differently across land cover classes rather than uniformly, providing practical, class-specific guidance for sensor selection in operational land cover mapping and improving the interpretability of machine learning-based mapping workflows. Full article

Precise geometric registration between high-resolution satellite imagery and aerial orthophotos is essential for generating high-definition (HD) maps that support autonomous vehicle navigation. This study presents a comprehensive evaluation of multi-spectral band performance for image registration between KOMPSAT-3A satellite imagery (0.55 m resolution) and VWorld aerial orthophotos (0.25 m resolution) across seven patch size configurations. Five feature-matching approaches were systematically compared: LightGlue with CLAHE preprocessing, edge-based FFT methods (with and without CLAHE), and SIFT-based methods (with and without CLAHE). Two additional detector-free deep matchers, LoFTR and RoMa, were further integrated into the same pipeline for comparison. The experimental results reveal significant variations in registration accuracy across spectral bands, with the panchromatic-derived products (SPECPAN and EMPPAN) and luminance composite BT601 image demonstrating superior stability compared to individual visible and NIR bands. LightGlue achieved consistently high inlier counts (averaging 1100+ matched points) across all spectral bands and patch configurations, while SIFT with CLAHE preprocessing yielded the lowest matching RMSE (averaging 1.55 pixels). Among all matchers, the detector-free methods produced the densest and most stable correspondences, with LoFTR giving the best transformation stability, whereas edge-based methods were markedly less stable. However, an independent assessment against network GNSS check points showed a registration accuracy of approximately 2.8 m that was statistically similar across all matchers, indicating that matcher selection mainly affects correspondence density and transformation stability rather than independent geodetic accuracy. The achieved meter-level accuracy is suitable for HD map preprocessing and candidate GCP generation rather than final lane-level mapping, and the reported guidance is specific to the tested KOMPSAT-3A/VWorld setting. Full article

Post-monsoon open-field stubble burning in northwestern (NW) India—a key agricultural region known as the “breadbasket”—is a longstanding practice used to clear fields. Satellite observations spanning over two decades have revealed significant upward trends in crop production, vegetative greenness, and the frequency of post-harvest fires, with this last contributing to hazardous air quality during the peak burning season (mid-October to mid-November). Since 2022, thermal anomaly data from Aqua-MODIS and SNPP-VIIRS sensors have shown a sharp decline in reported fire events—an observation that contrasts starkly with the concurrent rise in regional aerosol loading detected from space. This apparent discrepancy became particularly pronounced in 2024–2025, prompting a closer examination using high-temporal-resolution imagery from the Advanced Meteorological Imager (AMI) on the geostationary satellite GEO-KOMPSAT-2A. These observations revealed a clear spike in fire-related signals occurring around and after 4:00 p.m. local time, i.e., outside the typical noon to 2:00 p.m. detection window of the MODIS and VIIRS. A fire detection algorithm exploiting the fire-sensitive shortwave-infrared 3.8 μm signal and its contrast to 11.2 μm infrared observations is designed to adopt AMI observations and applied to its multi-year observations (2019–2025). The resulting fire dataset unambiguously shows a gradual shift in stubble burning activity toward the late afternoon hours beginning in 2022 which is underreported by polar-orbiting satellites. The orbital drift of NASA’s MODIS sensor on the Aqua platform allows detection of some of the gradually shifting fires during afternoon hours, but the MODIS still misses a large number of fires occurring around and after 4 p.m. The AMI’s relatively coarse spatial resolution (~4 km), a consequence of its slant viewing geometry over NW India, imposes inherent limitations on quantifying the full extent of fire occurrences. The operational air quality forecasting models currently assimilate satellite fire detections predominantly captured during early afternoon overpasses of the MODIS and VIIRS. The temporal shift in fire activity complicates such forecast, leading to a substantial underestimation of emissions. Intense stubble burning and the resulting air pollution highlight the need for effective crop residue management practices for mitigating the frequency of open biomass burning and thereby reducing episodic degradation of air quality and its associated public health and economic impacts. Full article

The Advanced Meteorological Imager (AMI) on GEO-KOMPSAT-2A (GK-2A) lacks a 2.1 μm shortwave infrared channel, precluding the dark target surface reflectance estimation that other geostationary aerosol retrievals rely on. We propose an improved land aerosol optical depth (AOD) retrieval in which background surface reflectance (BSR) is derived entirely from pixel-level bidirectional reflectance distribution function (BRDF) inversion using the scaled Ross-Thick Li-Sparse (sRTLS) kernel model fitted to geostationary time-series observations. Unlike existing approaches, the algorithm inverts the BRDF independently at each retrieval channel without relying on spectral reflectance relationships or external surface reflectance products; it assumes a low-background AOD during an initial accumulation period and then iteratively refines both BRDF coefficients and AOD. Two aerosol models—generic and dust—are supported, with a geographic dust-zone mask activating two-model selection during spring. Validation against 74 Aerosol Robotic Network sites over 2023 yields R = 0.86, RMSE = 0.15, and bias = −0.02, compared with R = 0.59, RMSE = 0.25, and bias = −0.04 for the National Meteorological Satellite Center (NMSC) GK-2A AOD product. The largest improvements appear at AOD ≤ 0.1 (bias: +0.03 versus +0.11) and AOD > 0.8 (bias: −0.12 versus −0.85). The full March–May (MAM) evaluation yields bias = −0.06 across all 74 sites. As a separate parallel retrieval restricted to matchups inside the geographic dust-zone mask, the proposed algorithm (dust model included) gives bias = −0.03, which worsens to −0.11 when only the generic model is applied—nearly a fourfold increase. A comparison against Himawari-9/Advanced Himawari Imager (AHI)—a co-located geostationary sensor carrying a 2.3 μm shortwave infrared (SWIR) channel—shows that the proposed algorithm (R = 0.897) outperforms Himawari-9/AHI (R = 0.855) across all metrics, demonstrating competitive accuracy without relying on a SWIR channel. Full article

In this study, a post-processing algorithm was developed to mitigate the over-detection tendency of the Geo-KOMPSAT-2A fog detection algorithm (GK2A_FDA) by integrating surface observations, facilitated by the recent availability of high-resolution gridded surface analysis data. The method was optimized for six sub-algorithms (inland/coastal × daytime/nighttime/twilight) using an interpretable decision tree model with data from 2021 to 2023. The $RH$ (relative humidity) and $\mathsf{\Delta }FTs$ (clear-sky background minus fog-top brightness temperature) step defines detection boundaries in a two-dimensional decision space using joint false alarm-to-hit ratio and hit count distributions to effectively remove false-alarm-dominated regions with minimal impact on the probability of detection (POD). The post-processing steps were sequenced according to independently quantified accuracy gains ($RH$ and $\mathsf{\Delta }FTs$ >> $T_a$ > wind speed > solar zenith angle), with thresholds conservatively derived and seasonally optimized for South Korea. Following post-processing, the POD decreased only slightly (0.08–0.27%), while the false alarm ratio (FAR) and bias were reduced by 5.13–13.68% and 16.13–52.61%, respectively. Improvements were more pronounced during drier seasons than wet seasons; however, the residual high daytime bias (3.348–5.319) indicated the need for further GK2A_FDA refinement. This study demonstrated that integrating satellite and surface observations could effectively address the limitations of satellite-based fog detection. Full article

This study evaluates the quality of ionosphere-corrected bending angle products from 12 satellite radio occultation (RO) missions, with data provided by the ROM SAF and CDAAC data centers. The missions include MetOp-B/C, Sentinel-6, Spire, COSMIC-2, KOMPSAT-5, and TerraSAR-X. The assessment focuses on bending angle quality control (QC), bias and noise characteristics at 65–80 km altitude, and statistical errors, with ERA5 data used as the reference. For quality control, Spire products achieved the highest pass rate, exceeding 99%. Products from the two MetOp satellites and Sentinel-6 exhibited pass rates above approximately 90%. The COSMIC-2 series had a pass rate of ~81%, while KOMPSAT-5 and TerraSAR-X had pass rates of 62% and 68%, respectively. Concerning bending angle biases, slightly larger biases were observed in MetOp setting events. Biases from other missions were mostly within the range of 0–0.05 μrad. Regarding noises, Sentinel-6 recorded the smallest bending angle noise (0.87 μrad), whereas TerraSAR-X (2.3 μrad) and KOMPSAT-5 (1.9 μrad) showed the largest noise magnitudes. Systematic differences in bending angles from all 12 RO missions were generally consistent below 60 km, while their standard deviations show good consistency below 35 km. In the middle stratosphere (35–50 km), MetOp-B/C and Sentinel-6 displayed the smallest standard deviations. Spire values are 1–2% larger, COSMIC-2 values 5–10% larger, and TerraSAR-X values the largest. Since ERA5 data also contain inherent uncertainties, particularly above 60 km, the findings of this study can only serve as a preliminary reference for users applying these datasets in weather and climate research. Future work will investigate the detailed causes of discrepancies among different datasets at high altitudes. Full article

The spatial quality (RER, FWHM, MTF50, and MTFA) of an on-orbit high-resolution Earth observation satellite was statistically assessed using edges extracted from natural targets in urban imagery, rather than from a ground-based artificial edge target (Edge target). A total of 8414 edge patches were automatically extracted from Natural Targets (NT) in 24 images acquired over 11 cities worldwide by KOMPSAT-3A. The spatial quality results derived from NT edges were compared with those obtained from an Edge target in terms of accuracy and precision. The averaged spatial quality values were nearly identical between the two approaches. However, the coefficient of variation (CV), used to quantify precision, was approximately two times higher for NT edges. Despite this lower precision, NT-based assessment provides substantial operational advantages. While 549 Edge target acquisitions over eight years yielded only 1098 usable edge images, a single acquisition over Las Vegas produced 366 NT edge images. These findings indicate that statistically reliable spatial quality estimates with comparable accuracy can be achieved within a single day using abundant NT edges, without the need to construct and maintain dedicated Edge targets. Full article

In this study, continuous near-real-time measurements of greenhouse gases (GHGs), particularly carbon dioxide (CO₂) and methane (CH₄), and aerosol optical depth (AOD) were conducted at the Socheongcho Ocean Research Station (SORS) from January 2021 to April 2022. Specifically, AOD products retrieved from the Geo-KOMPSAT-2B sensors—Geostationary Environment Monitoring Spectrometer and Geostationary Ocean Color Imager II—were compared and validated against ground-based Aerosol Robotic Network (AERONET) observations. Both satellite products exhibited overall good agreement with AERONET AOD data and showed low bias. The GHG measurements based on cavity ring-down spectroscopy indicated that CO₂ reached its highest seasonal mean in the spring of 2022, while CH₄ attained its maximum during the wet summer of 2022. Temperature, relative humidity, and evaporation were closely associated with AOD variability during the dry summer period, while elevated temperatures may have contributed to enhanced photochemical activity and modulation of CH₄ concentrations. In the cold season, concurrent increases in GHGs and combustion-related pollutants (PM_2.5, CO, and black carbon) were observed, suggesting reduced oxidation capacity under stable atmospheric conditions. Overall, these findings underscore the potential value of integrating satellite and in situ observations to better characterize GHG–aerosol interactions and support emission mitigation strategies in the Northeast Asian marine environment. Full article

Remote sensing images are often degraded by atmospheric effects, low illumination, and off-nadir viewing, which reduces the segmentation performance of deep models. KOMPSAT (Korea Multi-Purpose Satellite) imagery suffers from quality degradation because the Korean Peninsula is surrounded by sea on three sides and is subject to frequent weather and atmospheric variations. In practice, operators apply heuristic image enhancement techniques by hand, but these approaches are labor-intensive and inconsistent. To address this issue, we have proposed Adaptive Remote Sensing Image Enhancement (ARSIE), an automated reinforcement learning–based framework that improves segmentation performance on degraded KOMPSAT imagery. ARSIE takes only an existing segmentation network and training data as input, and learns, for each image, a sequence of enhancement operations selected from a filter pool. The policy network uses intermediate feature maps from the segmentation model to choose the next operation, ensuring that enhancement decisions directly support downstream segmentation performance. Experimental results show that ARSIE automatically discovers image-specific enhancement combinations and consistently improves segmentation accuracy on degraded KOMPSAT imagery. We demonstrate that ARSIE has the potential to be extended to improving the quality of other satellite imagery. Full article

Multispectral (MS) satellite imagery provides rich spectral information for surface and atmospheric interpretation, yet its spatial resolution is often limited by sensor design. In this study, we propose a Transformer-based MS super-resolution framework that uses high-resolution panchromatic (PAN) imagery to supply complementary spatial detail cues for MS reconstruction and explicitly separates spatial enhancement from spectral preservation. In the spatial branch, PAN features are aligned to the MS grid via Pixel-Unshuffle and encoded with shifted-window self-attention to capture long-range spatial dependencies efficiently. In the spectral branch, spectral self-attention treats bands as tokens to learn inter-band correlations and maintain spectral consistency. The two representations are fused through channel concatenation and a 1 × 1 convolutional module, followed by a reconstruction head that upsamples the fused features to generate high-resolution MS outputs. For training, low-resolution MS inputs are synthesized from KOMPSAT-3A MS imagery using a degradation pipeline that combines modulation transfer function-based blur, downsampling, and additive Gaussian noise; the operation order is randomly permuted to emulate diverse acquisition conditions. In addition, Bayesian optimization is employed to explore network configurations through jointly considering the normalized mean absolute error and inference time. Experiments demonstrate that the proposed approach attains 46.23 dB PSNR, 0.9735 SSIM, and 3.12 ERGAS with approximately 167.4 K parameters, achieving a high restoration quality and computational efficiency across diverse degradation settings. Full article