Article
Geochemistry & Geophysics
Yuan Li, Qizhi Xu, Zhaofeng He, Wei Li
Summary: This paper discusses the challenges of using infrared remote sensing images in real-world applications, including time-consuming preprocessing, decreased contrast, and difficulty in discriminating low-resolution targets. To address these challenges, a progressive task-based universal network is proposed, which includes a stripe denoising component, an object detection component, and a feedback loss adjustment mechanism. Experimental results demonstrate that the proposed method outperforms other state-of-the-art methods in infrared image ship detection.
IEEE TRANSACTIONS ON GEOSCIENCE AND REMOTE SENSING
(2023)
Article
Environmental Sciences
Haopeng Zhang, Xingyu Zhang, Gang Meng, Chen Guo, Zhiguo Jiang
Summary: Monitoring and identifying ships in remote sensing images is crucial for various applications, but ship detection remains challenging due to small size and complex backgrounds. In this paper, we propose a few-shot multi-class ship detection algorithm with attention feature map and multi-relation detector (AFMR) to address this problem. The proposed method achieves good detection performance on publicly available datasets and a self-constructed dataset.
Article
Environmental Sciences
Liyuan Li, Jianing Yu, Fansheng Chen
Summary: The article introduces the construction and utilization of a thermal infrared ship dataset, as well as the corresponding algorithm analysis and optimization, aiming to promote research and application of all-day ship detection.
Article
Environmental Sciences
Jun Zhang, Ruofei Huang, Yan Li, Bin Pan
Summary: This paper proposes a novel Oriented Ship detection method based on an intersecting Circle and Deformable region of interest (OSCD-Net) to address the challenges of ship detection, including the large length-width ratio and arbitrary direction of ships. OSCD-Net consists of two modules, ICR-head and DRoI, which perform detection and alignment to achieve promising performance on public remote sensing image datasets.
Article
Environmental Sciences
Zhenqing Wang, Yi Zhou, Futao Wang, Shixin Wang, Zhiyu Xu
Summary: The importance of ship detection in optical remote sensing images was highlighted, and a new fully convolutional neural network, SDGH-Net, based on Gaussian heatmap regression was proposed to improve ship recall rates without the need for non-maximum suppression. The method showed promising results in terms of F-measure and outperformed other comparison methods.
Article
Chemistry, Analytical
Huibin Li, Wei Guo, Guowen Lu, Yun Shi
Summary: This study addresses the issue of current lightweight detection models being ineffective in detecting multi-type occlusion targets during fruit picking. By introducing a multi-type occlusion apple dataset and a data balance augmentation method, the study shows significant improvement in the average detection precision of popular lightweight object detection models. The proposed augmentation method demonstrates great potential for future orchard applications in different fruit detection missions.
Article
Geochemistry & Geophysics
Zhida Ren, Yongqiang Tang, Zewen He, Lei Tian, Yang Yang, Wensheng Zhang
Summary: In this article, the authors propose using saliency information to aid ship detection, addressing the challenges of background confusion and sparse distribution of positive samples. The authors devise novel modules and a new sampling strategy to enhance ship detection in complex environments and increase the number of positive samples. Experimental results demonstrate the superiority of the proposed method over previous approaches.
IEEE TRANSACTIONS ON GEOSCIENCE AND REMOTE SENSING
(2022)
Article
Environmental Sciences
Bokun He, Qingyi Zhang, Ming Tong, Chu He
Summary: Recently, object detection in natural images has achieved breakthrough, while oriented ship detection in remote sensing imagery remains challenging. In this paper, an oriented ship detector based on pairwise branch detection head and adaptive SAR feature enhancement is proposed, which improves ship detection by addressing limitations including uncertain ship orientation, unspecific features, and SAR speckle interference.
Article
Remote Sensing
Zhenrong Zhuang, Wenzao Shi, Wenting Sun, Pengyu Wen, Lei Wang, Weiqi Yang, Tian Li
Summary: Change detection in remote sensing images has a significant impact on various applications. Recent advances have been made in change detection methods for different ground objects, but there are still limitations in feature recognition, resulting in unclear boundaries and a need for improved accuracy. To address these issues, the use of HRNet and new data augmentation methods are introduced to enhance accuracy. Additionally, the integration of CSWin and HRNet models improves performance, and a feature fusion network named A-FPN is designed to enhance perception of ground objects at different scales.
INTERNATIONAL JOURNAL OF REMOTE SENSING
(2023)
Article
Engineering, Ocean
Runyu Song, Tieshan Li, Taoying Li
Summary: Ship detection in marine remote sensing images is improved by a hybrid deep learning framework, which includes image dehazing, image enhancement, and an improved YOLO-v5s network. Experimental results demonstrate the superior ship detection precision of the proposed framework.
APPLIED OCEAN RESEARCH
(2023)
Article
Engineering, Marine
Pan Wang, Jianzhong Liu, Yinbao Zhang, Zhiyang Zhi, Zhijian Cai, Nannan Song
Summary: The proposed lightweight network efficiently detects and discriminates the directions of cargo ships in remote sensing images for autonomous operation on spaceborne platforms.
JOURNAL OF MARINE SCIENCE AND ENGINEERING
(2021)
Article
Engineering, Electrical & Electronic
Mingming Zhu, Guoping Hu, Hao Zhou, Shiqiang Wang, Yule Zhang, Shijie Yue, Yu Bai, Kexin Zang
Summary: This study proposes an arbitrary-oriented ship detection method based on RetinaNet, which achieves better detection accuracy through rotated RetinaNet, refined network, and improved loss function. Experimental results demonstrate that the proposed method outperforms other state-of-the-art methods on a new dataset.
IEEE JOURNAL OF SELECTED TOPICS IN APPLIED EARTH OBSERVATIONS AND REMOTE SENSING
(2021)
Article
Geochemistry & Geophysics
Weichang Zhang, Rui Zhang, Guoqing Wang, Wei Li, Xun Liu, Yang Yang, Die Hu
Summary: Automatic detection and localization of objects in remote sensing images are important for remote sensing systems. However, existing frameworks usually suffer from poor performance due to a lack of large-scale training datasets. To address this, a novel sensor-related image synthesis framework called RS-ISP is developed. It introduces designs to ensure distribution consistency between generated and real images, resulting in improved ship detection performance.
IEEE TRANSACTIONS ON GEOSCIENCE AND REMOTE SENSING
(2023)
Article
Engineering, Electrical & Electronic
Yangfan Li, Chunjiang Bian, Hongzhen Chen
Summary: Ship detection has been a hot topic in recent years, with various applications in military and civilian fields. This study proposes a dynamic soft label assignment method for arbitrary-oriented ship detection. The method introduces a novel anchor quality score function, a dynamic anchor quality score threshold, and a soft label assignment strategy to improve detection performance. Experimental results demonstrate the effectiveness of this method on different datasets.
IEEE JOURNAL OF SELECTED TOPICS IN APPLIED EARTH OBSERVATIONS AND REMOTE SENSING
(2023)
Article
Environmental Sciences
Ling Jian, Zhiqi Pu, Lili Zhu, Tiancan Yao, Xijun Liang
Summary: This study improves the performance of marine ship detection using self-supervised learning. By designing a CutPaste self-supervised task in multiple ways and training a feature representation network using clean marine surface images with no ships, the object detection model is enhanced to achieve better detection accuracy.
Article
Chemistry, Analytical
Linyi Jiang, Xiaoyan Li, Liyuan Li, Lin Yang, Lan Yang, Zhuoyue Hu, Fansheng Chen
Summary: Traditional methods of on-orbit geometric calibration using ground control points or stars as references for geostationary cameras face challenges in collecting high-precision GCPs due to cloud cover and extraction algorithms. In response, a novel approach using the relative motion of stars is proposed to improve positioning accuracy, overcoming the limitations of cloud cover and the number of observed stars.
