检索结果-内蒙古大学图书馆

IEEE/CVF conference on Computer Vision and pattern recognition (CVPR)

作者： Gao, Jianhao Li, Jie Su, Xin Jiang, Menghui Yuan, Qiangqiang Wuhan Univ Wuhan Peoples R China

ISBN: (数字)9781665487399

ISBN: (纸本)9781665487399

Pansharpening, whose aim is to acquire high resolution multispectral data (HRMS) by the fusion of low resolution multispectral data (LRMS) and panchromatic data (PAN), is a specific mission of spatial-spectral fusion in remote sensing field. In recent years, deep learning methods have proved the most feasible methods for pansharpening task. However, these deep learning methods have difficulty in training in an unsupervised manner and become useless when it comes to the condition where no training dataset is available. In this paper, we propose a universal algorithm called deep image interpolation for pansharpening task. The main idea is achieving high-quality fusion results by interpolating two low-quality multispectral images in a deep neural network. We apply it to two conditions: 1) unsupervised training a network when there are enough datasets;2) directly optimizing the fusion result where no training datasets are available. Simulation and real-data experiments are conducted on various kinds of satellite data. Quantitative and qualitative evaluation results illustrate that the proposed method outperforms traditional pansharpening methods and even catch up with those supervised methods to some extent.

关键词： Training Deep learning Interpolation Satellites Neural networks Data integration Pansharpening

来源：评论

学校读者我要写书评

暂无评论

Self-supervised Vision Transformers for Land-cover Segmentation and Classification

Self-supervised Vision Transformers for Land-cover Segmentat...

引用

IEEE/CVF conference on Computer Vision and pattern recognition (CVPR)

作者： Scheibenreif, Linus Hanna, Joelle Mommert, Michael Borth, Damian Univ St Gallen AIML Lab Sch Comp Sci Rosenbergstr 30 St Gallen Switzerland

ISBN: (纸本)9781665487399

Transformer models have recently approached or even surpassed the performance of ConvNets on computer vision tasks like classification and segmentation. To a large degree, these successes have been enabled by the use of large-scale labelled image datasets for supervised pre-training. This poses a significant challenge for the adaption of vision Transformers to domains where datasets with millions of labelled samples are not available. In this work, we bridge the gap between ConvNets and Transformers for Earth observation by self-supervised pre-training on large-scale unlabelled remote sensing data(1). We show that self-supervised pre-training yields latent task-agnostic representations that can be utilized for both land cover classification and segmentation tasks, where they significantly outperform the fully supervised baselines. Additionally, we find that subsequent fine-tuning of Transformers for specific downstream tasks performs on-par with commonly used ConvNet architectures. An ablation study further illustrates that the labelled dataset size can be reduced to one-tenth after self-supervised pre-training while still maintaining the performance of the fully supervised approach.

关键词： Training Earth Computer vision image segmentation Computational modeling conferences Transformers

来源：评论

学校读者我要写书评

暂无评论

Dualswin-Ynet: A Novel Bimodal Fusion Network for Ship Detection in remote sensing images 27th

Dualswin-Ynet: A Novel Bimodal Fusion Network for Ship Dete...

引用

27th International conference on pattern recognition, ICPR 2024

作者： Chen, Dongdong Ju, Rusheng Liu, Xiaoyang Liu, Jiyuan Zhang, Jun Qiu, Sihang College of Systems Engineering National University of Defense Technology No. 109 Deya Road Kaifu District Changsha China

ISBN: (纸本)9783031781681

Ship detection in remote sensing images has a wide range of research needs in both military and civilian applications. Traditional object detection algorithms rely solely on optical images as input, which are susceptible to various interferences such as cloud and fog occlusion, scene noise, etc., resulting in unsatisfactory detection accuracy. Based on the YOLO framework, this paper designs a ship detection model that integrates visible light and infrared images through dual-modal fusion, and explores various feature fusion methods. It achieves data-level fusion network SwinFuse-Ynet, feature-level fusion network E-Dualswin-Ynet and decision-level fusion network L-Dualswin-Ynet. Comprehensive validations are conducted on the multispectral satellite image datasets MMShip and VI-ship. The experimental results demonstrate that the three proposed networks effectively complement infrared feature information, reduce interference from complex scene factors, and improve detection accuracy. In particular, E-Dualswin-Ynet improved by introducing SwinTransformer and Wise-IoU can more fully amplify the complementarity between infrared and visible light modalities, obviously better than other single-modal detection methods. © The Author(s), under exclusive license to Springer Nature Switzerland AG 2025.

关键词： image enhancement

来源：评论

学校读者我要写书评

暂无评论

A novel sarnede method for real-time ship detection from synthetic aperture radar image

引用

MULTIMEDIA TOOLS AND APPLICATIONS 2022年第12期81卷 16921-16944页

作者： Raj, Anil J. Idicula, Sumam Mary Paul, Binu Cochin Univ Sci & Technol Dept Comp Sci Kochi Kerala India Cochin Univ Sci & Technol Sch Engn Div Elect Engn Kochi Kerala India

Deep learning-based ship detection from SAR data is one of the challenging problems in the remote sensing area. Also, SAR ship detection is precise object detection and pattern recognition task under the computer vision area. The main problems are false detection, primarily due to speckle presence and multi-scale SAR image availability. We propose a novel real-time system with a preprocessing technique exclusively for SAR ship detection to address this problem. The proposed SarNeDe preprocessing stage is specially designed using image processing techniques and lee filter to reduce the false prediction and improve the SAR image quality, which increases the detection accuracy because the lee filter alone could increase missed detections. The SarNeDe image is generated from raw SAR image and is given to a novel multi-scale lightweight deep learning model to predict all ships' positions. The proposed model has a feature merging & boosting network and three detection parts for detecting big, medium, & small ships. We experimented on the public SAR ship detection dataset (SSDD) and Dataset of Ship Detection for Deep Learning under Complex Backgrounds (SDCD) to validate the proposed method's feasibility. The experimental results indicated that our proposed method's ship detection accuracy is superior to other state-of-the-art ship detectors with reduced false detections.

关键词： Ship detection SAR Deep learning Convolution neural network Marine target detection

来源：评论

学校读者我要写书评

暂无评论

Multi-Scale Mixed Pixel Network for remote sensing image Super-Resolution

Multi-Scale Mixed Pixel Network for Remote Sensing Image Sup...

引用

pattern recognition and Artificial Intelligence (PRAI), International conference on

作者： He Yang Xun Ji Xu Wang Shijie Chen School of Marine Electrical Engineering Dalian Maritime University Dalian Liaoning China

ISBN: (数字)9798350350890

ISBN: (纸本)9798350350906

Deep learning-based image super-resolution (SR) technology has gained extensive attention in the remote sensing community, which aims to reconstruct the abundant details of target images. However, the practical application of existing deep learning-based SR methods is often limited by the trade-off between network performance and computational load. This paper presents a novel multi-scale mixed pixel network (MSMP-Net) to explore the precise and cost-effective SR reconstruction of remote sensing images. Specifically, a dual-path feature capture mechanism (DPFCM) architecture is devised that incorporates diverse dilated convolution operations to comprehensively capture hierarchical features across varying receptive fields. In addition, a triple-path high-resolution rebuilder (TPHRR) structure is designed, which uses a set of dense branches, residual branches and pyramid branches to effectively promote the transfer and reuse of features. Extensive experiments reveal that our proposed MSMP-Net outperforms other state-of-the-art SR methods in reconstruction effects.

关键词： Convolution Superresolution Poles and towers Learning (artificial intelligence) Computer architecture Streaming media Feature extraction pattern recognition remote sensing image reconstruction

来源：评论

学校读者我要写书评

暂无评论

High-speed and Low-latency 3D sensing with a Parallel-bus pattern 10

High-speed and Low-latency 3D Sensing with a Parallel-bus Pa...

引用

International conference on 3D Vision (3DV)

作者： Miyashita, Leo Tabata, Satoshi Ishikawa, Masatoshi Univ Tokyo 7-3-1 HongoBunkyo Ku Tokyo Japan Tokyo Univ Sci 1-3 KagurazakaShjinjuku Ku Tokyo Japan

ISBN: (纸本)9781665456708

High-speed 3D shape sensing is an essential technology for three-dimensional recognition and manipulation in dynamic scenes. However, conventional high-speed sensing methods mainly focus on the image capturing speed and the actual processing time to obtain a point cloud is not optimized in the measurement scheme and sensing pattern configuration. On the other hand, measurement latency is critical to respond in real-time and physically handle the dynamic scenes. This paper introduces a physically stereo-rectified projector-camera system for high-speed and low-latency 3D sensing with fast sequential memory access in the decoding process. Moreover, we configure a structured light pattern named "Parallel-bus pattern" with a De Bruijn torus and clock lines to maximize information density and robustly decode the pattern as data transfer in a parallel bus interface. We measured dynamically moving and deforming objects with the proposed system and evaluated the measurement performance. As a result, the developed 3D sensing system with the parallel-bus pattern achieved 27 K-points measurement at higher than 1000 fps with 0.336 ms latency and 0.838 mm accuracy on average.

关键词： De Bruijn torus high speed image processing low latency spatially encoded structured light

来源：评论

学校读者我要写书评

暂无评论

Self-Supervised Learning of remote sensing Scene Representations Using Contrastive Multiview Coding

Self-Supervised Learning of Remote Sensing Scene Representat...

引用

IEEE/CVF conference on Computer Vision and pattern recognition (CVPR)

作者： Stojnic, Vladan Risojevic, Vladimir Univ Banja Luka Fac Elect Engn Banja Luka Bosnia & Herceg

ISBN: (纸本)9781665448994

In recent years self-supervised learning has emerged as a promising candidate for unsupervised representation learning. In the visual domain its applications are mostly studied in the context of images of natural scenes. However, its applicability is especially interesting in specific areas, like remote sensing and medicine, where it is hard to obtain huge amounts of labeled data. In this work, we conduct an extensive analysis of the applicability of self-supervised learning in remote sensing image classification. We analyze the influence of the number and domain of images used for self-supervised pre-training on the performance on downstream tasks. We show that, for the downstream task of remote sensing image classification, using self-supervised pre-training on remote sensing images can give better results than using supervised pre-training on images of natural scenes. Besides, we also show that self-supervised pre-training can be easily extended to multispectral images producing even better results on our downstream tasks.

关键词： Visualization image segmentation image coding conferences Object detection Feature extraction pattern recognition

来源：评论

学校读者我要写书评

暂无评论

Self-Supervised Material and Texture Representation Learning for remote sensing Tasks

Self-Supervised Material and Texture Representation Learning...

引用

IEEE/CVF conference on Computer Vision and pattern recognition (CVPR)

作者： Akiva, Peri Purri, Matthew Leotta, Matthew Rutgers State Univ New Brunswick NJ 08854 USA Kitware Inc Clifton Pk NY USA

ISBN: (数字)9781665469463

ISBN: (纸本)9781665469463

Self-supervised learning aims to learn image feature representations without the usage of manually annotated labels. It is often used as a precursor step to obtain useful initial network weights which contribute to faster convergence and superior performance of downstream tasks. While self-supervision allows one to reduce the domain gap between supervised and unsupervised learning without the usage of labels, the self-supervised objective still requires a strong inductive bias to downstream tasks for effective transfer learning. In this work, we present our material and texture based self-supervision method named MATTER (MATerial and TExture Representation Learning), which is inspired by classical material and texture methods. Material and texture can effectively describe any surface, including its tactile properties, color, and specularity. By extension, effective representation of material and texture can describe other semantic classes strongly associated with said material and texture. MATTER leverages multi-temporal, spatially aligned remote sensing imagery over unchanged regions to learn invariance to illumination and viewing angle as a mechanism to achieve consistency of material and texture representation. We show that our self-supervision pretraining method allows for up to 24.22% and 6.33% performance increase in unsupervised and fine-tuned setups, and up to 76% faster convergence on change detection, land cover classification, and semantic segmentation tasks. Code and dataset: https : // github. com / periakiva/MATTER.

关键词： Representation learning Satellites Semantics Transfer learning Self-supervised learning Surface texture Task analysis

来源：评论

学校读者我要写书评

暂无评论

A fast bundle adjustment method based on track selection 4

A fast bundle adjustment method based on track selection

引用

4th International conference on Computer Vision and pattern Analysis, ICCPA 2024

作者： Liu, Changwei Xiong, Pingfan Yang, Luman Chen, Song Hu, Jinquan School of Remote Sensing and Information Engineering Wuhan University Hubei Wuhan430072 China School of Geodesy and Geomatics Wuhan University Hubei Wuhan430072 China Guangdon Shenzhen518000 China Wuhan Tianjihang Information Technology Co. Ltd. Hubei Wuhan430074 China

ISBN: (纸本)9781510682528

Bundle adjustment is the core of the Structure from Motion algorithm, and it is also a very time-consuming part, in which redundant observations and initial parameter values with large errors increase the time consumption of the algorithm. In order to improve the efficiency of bundle adjustment, we propose an track selection method based on uniformity, accuracy, coverage and connectivity criteria. Firstly, we divide the space of tracks into several 3D grids. Secondly, we start from the grid with the largest number of tracks, and eliminate the redundant tracks with low connectivity and low accuracy in each grid, while ensuring the connectivity and coverage of tracks. Finally, we use a variety of experimental data to verify this algorithm. The result shows that the algorithm can delete a large number of redundant tracks, and effectively improve the efficiency of the bundle adjustment method on the premise of ensuring the accuracy. When the track retention rate is 0.4, the efficiency of the bundle adjustment method is increased by about 2 times, and the corresponding precision loss value is 0.026. © 2024 SPIE.

关键词： Tunable filters 3D tracking image restoration Point clouds 3D image reconstruction image filtering image processing Pose estimation 3D modeling Atomic force microscopy

来源：评论

学校读者我要写书评

暂无评论

Approaching semi-supervised collaborative learning model for remote sensing image analysis

Approaching semi-supervised collaborative learning model for...

引用

2022 RIVF International conference on Computing and Communication Technologies, RIVF 2022

作者： Do, Viet Duc Mai, Dinh Sinh Ngo, Long Thanh Le Quy Don Technical University Dept. Information Technology Hanoi Viet Nam Institute of Techniques for Special Engineering Le Quy Don Technical University Hanoi Viet Nam

ISBN: (纸本)9781665461665

remote sensing image data often has many advantages in mapping, target recognition, object tracking, and so on. However, remote sensing image data often face problems such as big data, multiple dimensions, and time series. These things greatly affect the problem of remote sensing image analysis. In this paper, we extend the collaborative clustering algorithm for remote sensing image analysis by using a fuzzy semi-supervised technique. The proposed algorithm applies in the case of very few labeled data. In particular, the collaborative learning model can help reduce centralized computation at one site and instead process decentralized data at different sites. Thereby, it is possible to optimize the calculation ability with big data problems such as remote sensing image data. Moreover, by using a semi-supervised technique, it is possible to take advantage of a supervised clustering technique with a small amount of labeled data to improve the accuracy of clustering the results. Experiments on some actual remote sensing image data show that the proposed method gives significantly better results than the previously proposed methods. This result shows the potential of collaborative clustering in big data processing and improves the clustering quality by using semi-supervised fuzzy clustering with a small amount of labeled data. © 2022 IEEE.

关键词： Fuzzy clustering

来源：评论

学校读者我要写书评

暂无评论

建议与咨询留下您的常用邮箱和电话号码，以便我们向您反馈解决方案和替代方法

时间限定

文献类型

馆藏选择

核心期刊

语言

文献类型

帮助

文字说明：

检索规则说明：

检索范例：

分类表

所选分类

限定检索结果

文献类型

馆藏范围

日期分布

学科分类号

主题

机构

作者

语言

请选择保存的检索档案：

请选择收藏分类：

通借通还

建议与咨询 留下您的常用邮箱和电话号码，以便我们向您反馈解决方案和替代方法

时间限定

文献类型

馆藏选择

核心期刊

语言

文献类型

帮助

文字说明：

检索规则说明：

检索范例：

分类表

所选分类

限定检索结果

文献类型

馆藏范围

日期分布

学科分类号

主题

机构

作者

语言

请选择保存的检索档案： 新增检索档案 确定 取消

请选择收藏分类： 新增自定义分类 确定 取消

通借通还

建议与咨询留下您的常用邮箱和电话号码，以便我们向您反馈解决方案和替代方法

请选择保存的检索档案：

请选择收藏分类：