Article
Geochemistry & Geophysics
Runmin Dong, Lixian Zhang, Haohuan Fu
Summary: This research explores the use of reference-based super-resolution methods in remote sensing images, using texture information from high-resolution reference images to reconstruct details in low-resolution images. A novel end-to-end reference-based remote sensing GAN is proposed, which extracts features from reference images and aligns them with low-resolution features to transfer texture information to reconstructed high-resolution images. The proposed method shows robustness and outperforms existing methods in both quantitative evaluation and visual results.
IEEE TRANSACTIONS ON GEOSCIENCE AND REMOTE SENSING
(2022)
Article
Environmental Sciences
Pengcheng Zheng, Jianan Jiang, Yan Zhang, Chengxiao Zeng, Chuanchuan Qin, Zhenghao Li
Summary: This paper proposes a novel remote-sensing image super-resolution method called CGC-Net, which embeds context information and object priors from remote-sensing images into deep learning super-resolution models to achieve better image reconstruction performance.
Article
Environmental Sciences
Jiaao Li, Qunbo Lv, Wenjian Zhang, Baoyu Zhu, Guiyu Zhang, Zheng Tan
Summary: This study proposes a Multi-Attention Multi-Image Super-Resolution Transformer (MAST) that improves the application of multi-scale information and the modeling of the attention mechanism in multi-image super-resolution tasks. The proposed model achieves a superior restoration effect by better extracting and utilizing non-redundant information, balancing the global structure and local details of the image. Comparative experiments show notable enhancements in cPSNR, with improvements of 0.91 dB and 0.81 dB observed in the NIR and RED bands of the PROBA-V dataset, respectively, compared to existing state-of-the-art methods.
Review
Environmental Sciences
Xuan Wang, Jinglei Yi, Jian Guo, Yongchao Song, Jun Lyu, Jindong Xu, Weiqing Yan, Jindong Zhao, Qing Cai, Haigen Min
Summary: This paper provides a comprehensive overview and analysis of deep-learning-based image super-resolution methods for remote sensing images. It introduces the research background and details, presents important works and applications, and points out existing problems and future directions.
Article
Computer Science, Artificial Intelligence
Yi Xiao, Qiangqiang Yuan, Kui Jiang, Jiang He, Yuan Wang, Liangpei Zhang
Summary: In recent years, single image super-resolution (SR) has attracted significant attention in the remote sensing area, with numerous methods making remarkable progress in this field. However, most of these methods assume a fixed known degradation process, limiting their performance when faced with real-world distribution deviations. To address this, blind image super-resolution for multiple and unknown degradations has been explored. This paper proposes a self-supervised degradation-guided adaptive network to bridge the domain gap between simulation and reality and achieves superior results compared to state-of-the-art methods.
INFORMATION FUSION
(2023)
Article
Computer Science, Artificial Intelligence
Yi Wang, Syed Muhammad Arsalan Bashir, Mahrukh Khan, Qudrat Ullah, Rui Wang, Yilin Song, Zhe Guo, Yilong Niu
Summary: This paper reviews current datasets and deep learning-based object detection methods for remote sensing images. It proposes a large-scale Remote Sensing Super-resolution Object Detection (RSSOD) dataset and a novel Multi-class Cyclic super-resolution Generative adversarial network with Residual feature aggregation (MCGR) for benchmarking image super-resolution-based object detection. MCGR achieved state-of-the-art performance in image super-resolution and object detection.
EXPERT SYSTEMS WITH APPLICATIONS
(2022)
Article
Mathematics
Zhangzhao Cha, Dongmei Xu, Yi Tang, Zuo Jiang
Summary: Zero-shot super-resolution (ZSSR) has attracted much attention for its flexibility in various applications, but is ineffective for large-scale low-resolution image sets due to computational demands. To address this, we propose a novel meta-learning model that treats low-resolution images as ZSSR tasks and learns meta-knowledge. This approach reduces the computational burden and enhances the generalization capacity of ZSSR. Experimental results demonstrate its impressive performance and superiority over other state-of-the-art methods.
Article
Environmental Sciences
Jizhou Zhang, Tingfa Xu, Jianan Li, Shenwang Jiang, Yuhan Zhang
Summary: Limited resolution is a major obstacle to the application of remote sensing images (RSIs). This paper proposes a method to build a more realistic training dataset by modeling the degradation with blur kernels and imaging noises, and designs a novel residual balanced attention network (RBAN) to improve the super-resolution of images. Experimental results demonstrate that the proposed framework achieves state-of-the-art performance in quantitative evaluation and visual quality.
Article
Environmental Sciences
Haopeng Zhang, Cong Zhang, Fengying Xie, Zhiguo Jiang
Summary: Single image super-resolution (SISR) is an effective method to reconstruct a high-resolution (HR) image from a low-resolution (LR) input. We propose a closed-loop framework that utilizes the learning ability of the channel attention module and introduces information from real images. Experimental results show that our method performs superiorly in various training strategies.
Article
Chemistry, Multidisciplinary
Jian Guo, Mingkai Li, Qingjie Zhao, Qizhi Xu
Summary: This paper proposes a prediction-to-prediction super-resolution network under a multi-level supervision paradigm, which improves the performance of remote sensing image super-resolution reconstruction by increasing the number of supervisions and introducing flexible super-resolution components.
APPLIED SCIENCES-BASEL
(2023)
Article
Geochemistry & Geophysics
Sen Lei, Zhenwei Shi, Wenjing Mo
Summary: This research proposes a new framework for remote sensing image super-resolution by enhancing high-dimensional feature representation after upsampling layers using Transformers. Experimental results demonstrate that this method significantly improves super-resolution performance and outperforms several state-of-the-art methods.
IEEE TRANSACTIONS ON GEOSCIENCE AND REMOTE SENSING
(2022)
Article
Computer Science, Software Engineering
G. Rohith, Lakshmi Sutha Kumar
Summary: The article investigates the development of super-resolution algorithms from their inception to the latest technologies, emphasizing the importance and paradigm shifts of SR algorithms in the past 20 years. It proposes algorithms to solve SR problems in remote sensing applications, and demonstrates the effectiveness of deep learning-based SR algorithms in maintaining image clarity and enhancing spatial data through testing and analysis of publicly available images.
Article
Environmental Sciences
Yu Wang, Zhenfeng Shao, Tao Lu, Xiao Huang, Jiaming Wang, Xitong Chen, Haiyan Huang, Xiaolong Zuo
Summary: This paper proposes a remote sensing image super-resolution algorithm based on a multi-scale texture transfer network, which enhances the texture feature of reconstructed images by transferring texture information. The method adopts a multi-scale texture-matching strategy to obtain finer texture information.
Article
Environmental Sciences
Lixian Zhang, Runmin Dong, Shuai Yuan, Weijia Li, Juepeng Zheng, Haohuan Fu
Summary: This study proposes a method for super-resolution building extraction using relatively low-resolution images through a two-stage framework, achieving high-resolution building extraction. Experimental results demonstrate that the method performs well at different super-resolution ratios and outperforms eight other methods.
Article
Geochemistry & Geophysics
Hanlin Wu, Libao Zhang, Jie Ma
Summary: In this article, a saliency-guided feedback GAN (SG-FBGAN) is proposed to address the challenges posed by the versatile visual characteristics of different regions in remote sensing images (RSIs). The SG-FBGAN applies different reconstruction principles based on the saliency level of each region and uses feedback connections to improve expressivity while reducing parameters. A saliency-guided multidiscriminator is introduced to measure the visual perception quality of different areas and eliminate pseudotextures. Comprehensive evaluations and ablation studies validate the effectiveness of the proposed SG-FBGAN.
IEEE TRANSACTIONS ON GEOSCIENCE AND REMOTE SENSING
(2022)
Article
Engineering, Aerospace
Samet Aksoy, Aylin Yildirim, Taha Gorji, Nikou Hamzehpour, Aysegul Tanik, Elif Sertel
Summary: The study focuses on the importance of soil salinization detection using remote sensing techniques and machine-learning algorithms in arid and semi-arid regions. Among the three machine-learning algorithms analyzed, random forest algorithm demonstrated the most reliable spatial distribution of soil salinity classes in the selected study area, despite slightly better prediction results from classification and regression trees (CART) in some aspects.
ADVANCES IN SPACE RESEARCH
(2022)
Article
Environmental Sciences
Paria Ettehadi Osgouei, Gareth Roberts, Sinasi Kaya, Muhammad Bilal, Jadunandan Dash, Elif Sertel
Summary: Aerosol loading has significant impacts on radiative forcing, climate, and human health. This study evaluated and compared multiple satellite aerosol products in the Eastern Mediterranean and the Black Sea region, with VIIRS aerosol product showing the best performance over coastal areas.
ATMOSPHERIC ENVIRONMENT
(2022)
Article
Environmental Sciences
Raziye Hale Topaloglu, Gul Asli Aksu, Yusuf Alizade Govarchin Ghale, Elif Sertel
Summary: This study used GEOBIA and Landscape Metrics to create multi-temporal high-resolution LU/LC maps in a selected study region in the Istanbul metropolitan city of Turkey. The integration of open-source geospatial data improved the overall classification accuracy, and PCA was used to select landscape metrics and evaluate the results.
GEOCARTO INTERNATIONAL
(2022)
Article
Multidisciplinary Sciences
Paria Ettehadi Osgouei, Elif Sertel, M. Erdem Kabadayi
Summary: This study aims to determine the historical land use and land cover changes using different geospatial datasets from different time periods. The research proposes a method to prepare historical datasets and utilizes an object-oriented joint classification scheme to accurately map the spatio-temporal changes. The analysis shows diverging developments in the selected locations over a period of 162 years.
SCIENTIFIC REPORTS
(2022)
Article
Computer Science, Interdisciplinary Applications
Batuhan Sariturk, Dursun Zafer Seker, Ozan Ozturk, Bulent Bayram
Summary: This study investigates the performance evaluation of convolutional neural network architectures in building segmentation from high-resolution images. The results show that deeper architectures can provide better results even with limited data, and shallower architectures perform well with lower computational cost, making them useful for geographic applications.
EARTH SCIENCE INFORMATICS
(2022)
Article
Computer Science, Information Systems
Serdar Kizilkaya, Ugur Alganci, Elif Sertel
Summary: This study introduces the VHRShips dataset, which is a unique and rich ship dataset that can improve the scalability of ship detection and mapping applications. A deep learning-based multi-stage approach called HieD is proposed for ship type classification. The results show that HieD outperforms other methods in localization, recognition, and identification stages.
ISPRS INTERNATIONAL JOURNAL OF GEO-INFORMATION
(2022)
Article
Computer Science, Information Systems
Beytullah Sarica, Dursun Zafer Seker, Bulent Bayram
Summary: This study proposes a novel dense residual U-Net model that enhances automatic MS lesion segmentation using 3D MRI sequences. The model combines attention gate (AG), efficient channel attention (ECA), and Atrous Spatial Pyramid Pooling (ASPP) to achieve better results than other state-of-the-art methods. Validation on ISBI2015 and MSSEG2016 challenge datasets shows superior performance.
INTERNATIONAL JOURNAL OF MEDICAL INFORMATICS
(2023)
Article
Environmental Sciences
Kubra Bahsi, Beyza Ustaoglu, Samet Aksoy, Elif Sertel
Summary: This research aims to determine the amount of crop-based residue burning (CRB) in the Southeastern Anatolia Region of Turkiye, using Sentinel-2 images and Intergovernmental Panel on Climate Change standards. The analysis showed that CRB practices in 2019 released 14,444.307 Gg of Greenhouse Gases and 117.809 Gg of Particulate Matters. The results can improve national statistics and support agricultural decision-making processes.
GEOCARTO INTERNATIONAL
(2023)
Article
Computer Science, Artificial Intelligence
Peijuan Wang, Elif Sertel
Summary: With the advancement of artificial intelligence techniques and the launch of new satellites with video capturing capability, multi-frame super-resolution of remote sensing images has become a critical research topic. In this study, an attention-based Generative Adversarial Network (GAN) algorithm is proposed for multi-frame remote sensing image super-resolution. Several experiments were conducted, comparing the results of different models and the proposed approach using SpaceNet7 and Jilin-1 datasets.
KNOWLEDGE-BASED SYSTEMS
(2023)
Article
Geography, Physical
Tolga Bakirman, Bahadir Kulavuz, Bulent Bayram
Summary: This study aims to provide a holistic solution to monitor and protect cultural heritage from climate change, natural hazards, and anthropogenic effects. The efficiency of deep learning using low-cost unmanned aerial vehicles and camera images for the documentation and monitoring of cultural heritage is investigated. The proposed solution can aid in monitoring the protection of cultural heritage from climate change, natural disasters, and anthropogenic effects.
PHOTOGRAMMETRIC ENGINEERING AND REMOTE SENSING
(2023)
Article
Engineering, Geological
Bugrahan oezcihan, Levent Dogukan oezlue, Muemin Ilker Karakap, Halime Suermeli, Ugur Alganci, Elif Sertel
Summary: Satellite images are widely used in the production of geospatial information. Geometric correction is essential for image pre-processing to extract accurate locational information. This study performed geometric correction on satellite images obtained from Pleiades 1A (PHR) and SPOT-6 using empirical and physical models. Several experiments were conducted to investigate the effects of various parameters on the performance of the geometric correction procedure. The results showed that the model using RPC from data providers achieved lower RMSE values, providing better locational accuracy.
INTERNATIONAL JOURNAL OF ENGINEERING AND GEOSCIENCES
(2023)
Article
Remote Sensing
Haydar Akcay, Samet Aksoy, Sinasi Kaya, Elif Sertel, Jadu Dash
Summary: This study evaluated the performance of Sentinel-1 and Sentinel-2 data fusion for regional mapping of olive trees. Using the Izmir Province in Turkiye as a case study, the researchers successfully produced a regional-scale olive distribution map. The results showed that this method can be scaled up to the entire country and replicated elsewhere, providing a foundation for other scientific studies and effective management practices.
INTERNATIONAL JOURNAL OF REMOTE SENSING
(2023)
Proceedings Paper
Engineering, Aerospace
Ramil Safarov, Elif Sertel
Summary: This study used Remote Sensing and Geographic Information Systems to analyze the coastal changes around the Baku International Sea Trade Port over the past thirty years. The results showed that the most significant shoreline change occurred on the east side of the port, while the smallest change was observed 11.5 kilometers south of the port. Additionally, the study found that the land area had increased by 633 hectares due to coastline filling.
2023 10TH INTERNATIONAL CONFERENCE ON RECENT ADVANCES IN AIR AND SPACE TECHNOLOGIES, RAST
(2023)
Article
Computer Science, Software Engineering
Buket Bayram, Bahadir Kulavuz, Berkay Ertugrul, Bulent Bayram, Tolga Bakirman, Tuna Cakar, Metehan Dogan
Summary: Skin cancer, one of the most dangerous cancer types, requires early detection for successful recovery. This study demonstrates that deep learning-based image classification can aid doctors in accurately diagnosing skin lesions, achieving high accuracy rates.
BALTIC JOURNAL OF MODERN COMPUTING
(2022)
Article
Geochemistry & Geophysics
Cengiz Avci, Elif Sertel, Mustafa Erdem Kabadayi
Summary: The research aimed to propose an ideal architecture, encoder, and hyperparameter settings for historical road extraction, achieving the best results with the combination of Unet++ architecture and split-attention network (Timm-resnest200e) encoder, which can directly be used for inference of other historical maps and transfer learning.
IEEE GEOSCIENCE AND REMOTE SENSING LETTERS
(2022)