Jul 2026· The International Archives of the Photogrammetry, Remote Sensing and Spatial Information Sciences· 0 citations· 11 references
TL;DR
The results show that augmentation is most useful for single-acquisition small satellite imagery, where it improves robustness to lower image quality, indicating that performance depended more on scene conditions than on perturbation strength.
Abstract
Abstract. Road extraction from small satellite imagery is challenging because raw images often suffer from low signal-to-noise ratio (SNR), high radiometric variability, and reduced sharpness. In this work, we investigate whether noise and blur data augmentation during pretraining can improve robustness in such conditions. We use a two-stage transfer-learning framework in which a U-Net with a ResNet-50 encoder is pretrained on PlanetScope RGB imagery and fine-tuned on NEMO-HD imagery. During pretraining, we evaluate Gaussian, ISO-like, and Perlin noise, as well as Gaussian and motion blur, each at three severity levels. On the internal held-out test split, augmentation effects were modest, with the best strict IoU improving from 26.7% for the geometric-only baseline to 27.2%. However, evaluation on external full-scene NEMO-HD images showed clearer benefits. Augmentation-based models consistently improved road detection in raw imagery, mainly by increasing completeness and recall, while there was little or no systematic benefit in stacked imagery. No clear trend was observed across augmentation severity levels, indicating that performance depended more on scene conditions than on perturbation strength. The results show that augmentation is most useful for single-acquisition small satellite imagery, where it improves robustness to lower image quality.
Ground-to-Air (G2A) drone detection in medium- and long-wave infrared (MWIR/LWIR) imagery is challenging due to reduced texture information, sensor noise, weak thermal contrast, and the scarcity of annotated data. This work investigates a synthetic-first training strategy that combines synthetic scene generation with fine-tuning on real data. We show that synthetic data provides an effective basis for learning initial object representations, while real in-domain thermal imagery is still essential for reliable deployment. Even small amounts of real IR data substantially reduce domain gaps. Our experiments indicate that dataset alignment has a stronger impact on performance than model scale. Finally, our analysis of the dataset suggests that semantic alignment in feature space is the strongest predictor of model performance, while radiometric properties such as entropy and dynamic range also contribute to detection robustness. This work provides a foundation for combining synthetic and real IR data for effective G2A drone detection.
Tanel Liiv, Sander Soodla, Nzamba Bignoumba et al.· 0 citations
Detecting small targets in UAV aerial imagery is inherently difficult because these objects occupy only a small number of pixels and are highly susceptible to cluttered backgrounds, dense spatial arrangements, and pronounced scale variation. To address this problem, we propose CASA-Net (Context-Aware Small-object Adaptation Network), a context-aware detector built on a YOLOv26s baseline with three coordinated improvements: an Enhanced Small-Target-Aware Label Assignment mechanism for stronger supervision of tiny instances, a Multi-scale Feature Enhancement Module for richer contextual representation and spatial discrimination, and an aerial-specific augmentation pipeline for improved robustness to viewpoint, scale, and motion blur. Experiments on the VisDrone and RSOD benchmarks demonstrate that CASA-Net consistently outperforms the baseline and competing methods. On VisDrone, it achieves 47.0% mAP0.5 and 25.2% small-object mAP0.5, while on RSOD it reaches 78.3% mAP0.5. In addition, the model achieves real-time inference speeds above 100 FPS on both datasets using an RTX 3090, with 13.9 M parameters and 26.2 GFLOPs. Taken together, these findings show that CASA-Net is an accurate and efficient framework for UAV small-object detection through the joint improvement of supervision, feature representation, and data adaptation.
Kai Zheng, Yichen Zhong, Wenguang Song et al.· Remote Sensing· 0 citations
The upgraded version of the geometric correction module of the STORM processing chain can automatically orthorectify images from the NEMO-HD small satellite, which, like other small satellites, in principle has a lower signal-to-noise ratio (SNR) and higher radiometric variability.
Aleš Marsetič, P. Pehani, Nina Krašovec· The International Archives o...· 0 citations
Satellite imagery often suffers from limited spatial resolution and, in many cases, high acquisition costs. These factors restrict their use in applications such as urban monitoring, land management, and wildlife studies. This work proposes an AI-based super-resolution approach that leverages high resolution aerial imagery to train a Generative Adversarial Network. Specifically, the ESRGAN (Enhanced Super-Resolution Generative Adversarial Network) architecture is adapted and trained using aerial orthophotos, enabling the transfer of learned spatial representations to low-resolution satellite images. The trained model is evaluated on satellite image patches at 2 and 4 super-resolution scales. Performance is assessed using structural, perceptual, and chromatic metrics, including SSIMY, MS-SSIM, LPIPS and CIEDE2000. The results show clear improvements, with increased sharpness, enhanced edge definition, and consistent reconstruction of urban structures and terrain features. From a quantitative perspective, the 2 scale achieves the best overall metric values, while the 4 scale maintains stable and meaningful performance despite the higher reconstruction difficulty. These findings demonstrate the feasibility of transferring super-resolution capabilities from aerial images to satellite imagery, even in the presence of spectral and geometric differences between acquisition domains. Overall, this study provides a solid foundation for the development of low-cost, AI-driven satellite image super-resolution models and outlines future research directions focused on dataset expansion, domain adaptation strategies, and sensor-specific architectural improvements.
Magda Alexandra Trujillo-Jiménez, Francisco Iaconis, Debora Pollicelli et al.· IEEE Latin America Transacti...· 0 citations
Poor illumination obscures small, low-contrast defects in UAV bridge imagery, reducing the reliability and operational flexibility of automated inspection. This paper investigates whether degradation-aware image restoration can improve bridge damage detection under low-light conditions and transfer from synthetic degradations to real inspection scenes. We propose DaL- MoE, a detector-agnostic restoration front end trained with an ISP-aware low-light synthesis pipeline and equipped with degradation-aware guidance estimation and complementary experts for noise suppression, color adjustment, and structural-detail recovery. On paired synthetic data, DaL-MoE achieves 23.12 dB PSNR and 0.8482 SSIM, increasing YOLOv11m box mAP50 from 0.3097 to 0.4923 and mask mAP50 from 0.2281 to 0.3529. On real low-light UAV imagery without paired normal-light references, sim-to-real evaluation shows improved defect visibility and more complete detections than direct inference on raw low-light inputs. Future work will develop low-light-aware bridge damage detectors with stronger cross-scene generalization across bridge sites, imaging conditions, and illumination levels.
Hu Wang, Hong-Xu Pu, Zhi-ang Hu et al.· 0 citations
Small-object detection in unmanned aerial vehicle (UAV) imagery is challenging because targets are often tiny, densely distributed, and embedded in cluttered backgrounds, while edge platforms impose strict computational constraints. This paper presents a lightweight end-to-end detector based on RT-DETR to improve detection accuracy, localisation quality, and deployment efficiency. The proposed method includes three components: a detail-enhanced backbone for preserving fine-grained texture and edge cues, rotary positional encoding combined with a content-aware bidirectional feature pyramid for improved cross-scale alignment, and a difficulty-aware composite regression loss for more stable bounding-box optimisation. Experiments on VisDrone2019 and DOTA show that the proposed method improves mAP50 by 3.5% and 1.6%, respectively, over the RT-DETR baseline, while reducing the number of parameters by 28.6%. Ablation studies confirm the complementary contributions of the three components, and robustness experiments demonstrate reliable performance under challenging aerial conditions. Deployment on a Jetson Orin NX using a self-built UAV dataset further validates the practicality of the method for edge-side UAV monitoring.
Na Liu, Xiaoying Liao, Haotian Song et al.· 2026 IEEE 27th China Confere...· 0 citations
We use cookies to run the site and, with your consent, for analytics and to show ads.
See our Cookie Policy.
