检索结果-内蒙古大学图书馆

Temporal Optimisation of Satellite Image-Based Crop Mapping: A Comparison of Deep Time Series and Semi-Supervised Time Warping Strategies

引用

IET COMPUTER VISION 2025年第1期19卷

作者： Finnegan, Rosie Metcalfe, Joseph Sharifzadeh, Sara Caraffini, Fabio Xie, Xianghua Hornero, Alberto Synes, Nicholas W. Swansea Univ Dept Comp Sci Swansea Wales CSIC Spanish Natl Res Council IAS Inst Agr Sostenible Cordoba Spain Univ Melbourne Fac Engn & Informat Technol FEIT Melbourne Australia Agreena Copenhagen Denmark

This study presents a novel approach to crop mapping using remotely sensed satellite images. It addresses the significant classification modelling challenges, including (1) the requirements for extensive labelled data and (2) the complex optimisation problem for selection of appropriate temporal windows in the absence of prior knowledge of cultivation calendars. We compare the lightweight Dynamic Time Warping (DTW) classification method with the heavily supervised convolutional neural Network - Long Short-Term Memory (CNN-LSTM) using high-resolution multispectral optical satellite imagery (3 m/pixel). Our approach integrates effective practical preprocessing steps, including data augmentation and a data-driven optimisation strategy for the temporal window, even in the presence of numerous crop classes. Our findings demonstrate that DTW, despite its lower data demands, can match the performance of CNN-LSTM through our effective preprocessing steps while significantly improving runtime. These results demonstrate that both CNN-LSTM and DTW can achieve deployment-level accuracy and underscore the potential of DTW as a viable alternative to more resource-intensive models. The results also prove the effectiveness of temporal windowing for improving runtime and accuracy of a crop classification study, even with no prior knowledge of planting timeframes.

关键词： computer vision convolutional neural nets learning (artificial intelligence) remote sensing time series

来源：评论

学校读者我要写书评

暂无评论

Open-DS: An Open-Set Modulation Recognition Method for Low Probability of Interception Radar Emitters Based on Dictionary Similarity

引用

IET RADAR SONAR AND NAVIGATION 2025年第1期19卷

作者： Sun, Gangyin Chen, Shiwen Wu, Chaopeng Zhang, Li Fang, Haikun Natl Digital Switching Syst Engn & Technol Res Ctr Zhengzhou Peoples R China State Key Lab Math Engn & Adv Comp Zhengzhou Peoples R China

The recognition of radar emitters modulation in an open-set scenario presents a challenging task, particularly when identifying unknown modulation. This paper proposes a dictionary similarity based method for low intercept probability radar signal open-set modulation recognition (OMR), designed to address the unknown modulation in open-set scenarios. First, deep features of the input 1-D signal are extracted, and a random Fourier transform is applied to map the signal into a high-dimensional space, thereby converting the nonlinear feature optimisation problem into a linear optimisation problem. Next, an inter-class discreteness (ICD) module and an intra-class similarity (ICS) module are designed. Based on the Hilbert-Smith independence criterion, the correlation between features is quantified, and the quantitative values of ICD and ICS are used as loss functions to constrain the network's learning process. This approach effectively enhanced the representational power of the class dictionaries and significantly improved the model's overall performance. Experimental results demonstrate that the proposed strategy successfully extracts high-dimensional feature prototypes, achieving high accuracy in closed-set recognition while effectively performing open-set recognition tasks.

关键词： convolutional neural nets decorrelation electronic countermeasures LPI radar phase modulation radar radar waveforms signal classification time series

来源：评论

学校读者我要写书评

暂无评论

CSS-UNet: Convolution-State Space Enhanced UNet for Cardiac MRI Segmentation

引用

ELECTRONICS LETTERS 2025年第1期61卷

作者： Tong, Yaliang Liu, Kun He, Yuquan Yang, Ping Jilin Univ China Japan Union Hosp Dept Cardiol Changchun Jilin Peoples R China

Cardiac magnetic resonance imaging (CMR) is widely adopted in clinic for the assessment of cardiac anatomical structures and functions, and accurately segmenting left ventricular, myocardium and right ventricle from CMR plays an important role in clinical practice. convolutional neural networks (CNNs), for example, U-Net, have been widely used in CMR segmentation. However, current CNN-based models focus on extracting local features via convolution modules, which cannot well understand the long-range dependencies within images, leading to the sub-optimal solution for CMR segmentation task. Inspired by Mamba that efficiently captures global context information using state space model, we propose a novel CMR segmentation model, called CSS-UNet, that can capture both local features and global contexts simultaneously by fusing features from convolution block and visual state space block. Our new model follows the design of U-Net architecture that contains encoder and decoder with skip connections, where a proposed feature fusion module, called spatial-state space, is seamlessly integrated into our model. By using the spatial-state space module, the low-level and high-level features can be extracted and fused for capturing both global and local information, enhancing the capability of feature extraction of CSS-UNet. We evaluate our proposed model on two public CMR datasets, and the experimental results reveal that our proposed model outperforms the most widely-used UNet, demonstrating the effectiveness of our model.

关键词： convolutional neural nets image segmentation medical image processing magnetic resonance imaging

来源：评论

学校读者我要写书评

暂无评论

A structure-prior guided adaptive context selection network for remote sensing semantic segmentation

引用

ELECTRONICS LETTERS 2025年第1期61卷

作者： Xu, Shengjun Shen, Rui Liu, Erhu Ma, Zongfang Du, Miao Liu, Jun Zhan, Bohan Xian Univ Architecture & Technol Sch Informat & Control Engn Xian Peoples R China Xi An Jiao Tong Univ Sch Elect Engn Xian Peoples R China

In remote sensing image segmentation, recognizing buildings is challenging when the visual evidence from pixels is weak or when buildings belong to small, spatially structured objects. To address this issue, a structure-prior guided adaptive context selection network (SGACS-Net) is proposed for remote sensing semantic segmentation. The core is to use structure-prior knowledge to dynamically capture prior contextual information and higher-order object structural features, thereby improving the accuracy of remote sensing building segmentation. First, an adaptive context selection module is designed. By dynamically adjusting the spatial sensing field, this module effectively models the global long-range context information dependencies. It captures varying context information of buildings at different scales, thereby enhancing the network's ability to extract building feature representations. Second, a structure-prior guided variable loss function is proposed. It utilizes the structural features of building points, lines, and surface to identify key regions. By leveraging advanced structure-prior knowledge, it enhances the network's ability to express structural features. Experimental results on two datasets show that the proposed SGACS-Net outperforms other typical and state-of-the-art methods in terms of remote sensing semantic segmentation performance.

关键词： computer vision convolutional neural nets geophysical image processing

来源：评论

学校读者我要写书评

暂无评论

Design and evaluation of multidimensional CNN simulation framework against synchrophasor data spoofing attacks and correction for cyber-resiliency in power system

引用

IET GENERATION TRANSMISSION & DISTRIBUTION 2025年第1期19卷

作者： Thomas, Robert Nobili Britto Thirugnanasambandam, Venkatesh Senthur, Manikandan Rekshana, Manusri SASTRA Deemed be Univ Sch Elect & Elect Engn Thanjavur Tamil Nadu India

The increasing demand on synchronized measurements obtained from phasor measurement units (PMUs) has further increased potential threats, including data spoofing attacks. Such attacks on the synchronized measurements might greatly undermine the dynamic state estimation and important power system applications. This work proposes an integrated framework consisting of cyber physical testbed and multidimensional convolutional neural network (CNN) simulation framework to classify the data spoofing attacks on the generated real world synchrophasor data. This work uses a source authentication-based detection technique by extracting the spatial fingerprint information from PMU data. This information is provided as inputs to the multi-dimensional CNN (MD CNN) simulation framework. Finally, a robust correction algorithm using generative adversarial network (GAN) for reconstruction of spoofed PMU signal is proposed to eliminate bad measurements and enhance the integrity and reliability of synchrophasor data. Finally, the results of the proposed MD CNN classification framework and GAN based signal reconstruction algorithm are validated against the results obtained using traditional methods to validate the effectiveness of the proposed approach. Thus, the proposed framework aims at ensuring the cyber resilience of power grids to ensure its safe and reliable operation.

关键词： convolutional neural nets cyber-physical systems data protection signal reconstruction smart power grids WAMS/WAC/PMU

来源：评论

学校读者我要写书评

暂无评论

GDS-YOLO: A Rice Diseases Identification Model With Enhanced Feature Extraction Capability

引用

IET IMAGE PROCESSING 2025年第1期19卷

作者： Huang, Yourui Feng, Xi Han, Tao Song, Hongping Liu, Yuwen Bao, Meiping Anhui Univ Sci & Technol Sch Elect & Informat Engn Huainan Peoples R China Anhui Polytech Univ Wuhu Peoples R China Anhui Univ Sci & Technol Sch Artificial Intelligence Huainan Peoples R China

Accurate identification of rice diseases is a prerequisite for improving rice yield and quality. However, the rice diseases are complex, and the existing identification models have the problem of weak ability to extract rice disease features. To address this issue, this paper proposes a rice disease identification model with enhanced feature extraction capability, named GDS-YOLO. The proposed GDS-YOLO model improves the YOLOv8n model by introducing the GsConv module, the Dysample module, the spatial context-aware module (SCAM) and WIoU v3 loss functions. The GsConv module reduces the model's number of parameters and computational complexity. The Dysample module reduces the loss of the rice diseases feature during the extraction process. The SCAM module allows the model to ignore the influence of complex backgrounds and focus on extracting rice disease features. The WIoU v3 loss function optimises the regression box loss of rice disease features. Compared with the YOLOv8n model, the P and mAP50 of GDS-YOLO increased by 5.4% and 4.1%, respectively, whereas the number of parameters and GFLOPS decreased by 23% and 10.1%, respectively. The experimental results show that the model proposed in this paper reduces the model complexity to a certain extent and achieves good rice diseases identification results.

关键词： agricultural engineering convolution convolutional neural nets identification image processing image recognition

来源：评论

学校读者我要写书评

暂无评论

MCFNet: Research on small target detection of RGB-infrared under UAV perspective with multi-scale complementary feature fusion

引用

IET IMAGE PROCESSING 2025年第1期19卷

作者： Jing, Jing Hu, Jian Feng Zhao, Zuo Peng Liu, Ying Jiangsu Union Tech Inst Xuzhou Finance & Econ Branch Xuzhou Peoples R China China Univ Min & Technol Sch Comp Sci Xuzhou 221116 Jiangsu Peoples R China

The task of single-modal image target detection from a drone's perspective faces multiple challenges such as large image sizes, small and dense targets, insufficient lighting conditions, and hardware resource constraints, all of which affect the accuracy and real-time performance of algorithms. In response, a small target detection algorithm that fuses multi-scale complementary features, named MCFNet is proposed. Firstly, in order to independently extract the unimodal features of visible light and infrared images and effectively enhance the complementary information between images, a dual-stream backbone network with a terminal fusion mechanism is proposed. Secondly, the complementary feature information residual module is utilized to optimize the integration process of residual features. Subsequently, by designing a multi-scale feature enhancement module the network's capability to capture multi-scale features is enhanced. Finally, for targets of varying sizes, a lightweight Transformer feature extraction module is proposed to improve the detection accuracy of small targets from a drone's perspective. Test results on the drone-vehicle dataset show that this method achieved an average detection accuracy of 67.92%, while the detection accuracy on the self-constructed UAV-data dataset reached 96.4%. Additionally, a series of ablation experiments validated the effectiveness of the different modules.

关键词： computer vision convolution convolutional neural nets data handling feature extraction

来源：评论

学校读者我要写书评

暂无评论

RSTNet: A Spatio-Temporal Attention Framework for Human Action Recognition

引用

ELECTRONICS LETTERS 2025年第1期61卷

作者： Cui, Haifeng Hou, Zhihong Zhang, Tianyu Duan, Daxin Yao, Mingkai Liu, Taoran Shang, Mingwei Qu, Yang Wang, Yafei Wang, Hongbo Yao, Tianming Tian, Baofeng 964 Hosp Joint Logist Support Force Peoples Libera Trauma Surg Changchun Peoples R China Jilin Univ Coll Instrumentat & Elect Engn Key Lab Geophys Explorat Equipment Minist Educ Changchun Peoples R China

This paper introduces RSTNet, a neural network model based on spatio-temporal attention (STA), designed to improve the accuracy of human action recognition. The model uses heatmaps as input, employs 3D-ResNet as its backbone network, and incorporates STA modules and squeeze-and-excitation (SE) modules. Experiments on the UCF101 dataset demonstrate that RSTNet outperforms other classic methods in key metrics such as Top1 accuracy, Top5 accuracy and average accuracy. Ablation studies further validate the contribution of each module to the model's performance, proving the effectiveness of this approach in capturing spatio-temporal features and enhancing action recognition precision.

关键词： computer vision convolutional neural nets image classification image recognition learning (artificial intelligence)

来源：评论

学校读者我要写书评

暂无评论

Structure-Based Uncertainty Estimation for Source-Free Active Domain Adaptation

引用

IET Computer Vision 2025年第1期19卷

作者： Ouyang, Jihong Zhang, Zhengjie Meng, Qingyi Chi, Jinjin College of Computer Science and Technology Jilin University Changchun China Key Laboratory of Symbolic Computation and Knowledge Engineering of Ministry of Education Jilin University Changchun China

Active domain adaptation (active DA) provides an effective solution by selectively labelling a limited number of target samples to significantly enhance adaptation performance. However, existing active DA methods often struggle in real-world scenarios where, due to data privacy concerns, only a pre-trained source model is available, rather than the source samples. To address this issue, we propose a novel method called the structure-based uncertainty estimation model (SUEM) for source-free active domain adaptation (SFADA). To be specific, we introduce an innovative active sample selection strategy that combines both uncertainty and diversity sampling to identify the most informative samples. We assess the uncertainty in target samples using structure-wise probabilities and implement a diversity selection method to minimise redundancy. For the selected samples, we not only apply standard-supervised loss but also conduct interpolation consistency training to further explore the structural information of the target domain. Extensive experiments across four widely used datasets demonstrate that our method is comparable to or outperforms current UDA and active DA methods. © 2025 The Author(s). IET Computer Vision published by John Wiley & Sons Ltd on behalf of The Institution of Engineering and Technology.

关键词： computer vision convolutional neural nets image classification

来源：评论

学校读者我要写书评

暂无评论

Keynote: Small neural nets Are Beautiful: Enabling Embedded Systems with Small Deep-neural-Network Architectures 12

Keynote: Small Neural Nets Are Beautiful: Enabling Embedded ...

引用

International Conference on Hardware/Software Codesign and System Synthesis (CODES+ISSS)

作者： Iandola, Forrest Keutzer, Kurt DeepScale Mountain View CA 94043 USA Univ Calif Berkeley EECS Berkeley CA USA

ISBN: (纸本)9781450351850

Over the last five years Deep neural nets have offered more accurate solutions to many problems in speech recognition, and computer vision, and these solutions have surpassed a threshold of acceptability for many applications. As a result, Deep neural Networks have supplanted other approaches to solving problems in these areas, and enabled many new applications. While the design of Deep neural nets is still something of an art form, in our work we have found basic principles of design space exploration used to develop embedded microprocessor architectures to be highly applicable to the design of Deep neural Net architectures. In particular, we have used these design principles to create a novel Deep neural Net called SqueezeNet that requires only 480KB of storage for its model parameters. We have further integrated all these experiences to develop something of a playbook for creating small Deep neural nets for embedded systems.

关键词： Embedded computer vision Deep Learning Deep neural nets convolutional neural nets

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：