检索结果-内蒙古大学图书馆

UAV-supported intelligent truth discovery to achieve low-cost communications in mobile crowd sensing

Digital Communications and Networks 2024年第4期10卷 837-852页

作者： Jing Bai Jinsong Gui Guosheng Huang Shaobo Zhang Anfeng Liu School of Computer Science and Engineering Central South UniversityChangSha410083China School of Information Science and Engineering Hunan First Normal UniversityChangsha410205China School of Computer Science and Engineering Hunan University of Science and TechnologyXiangTan411100China

Unmanned and aerial systems as interactors among different system components for communications,have opened up great opportunities for truth data discovery in Mobile Crowd Sensing(MCS)which has not been properly solved in the *** this paper,an Unmanned Aerial Vehicles-supported Intelligent Truth Discovery(UAV-ITD)scheme is proposed to obtain truth data at low-cost communications for *** main innovations of the UAV-ITD scheme are as follows:(1)UAV-ITD scheme takes the first step in employing UAV joint Deep Matrix Factorization(DMF)to discover truth data based on the trust mechanism for an Information Elicitation Without Verification(IEWV)problem in MCS.(2)This paper introduces a truth data discovery scheme for the first time that only needs to collect a part of n data samples to infer the data of the entire network with high accuracy,which saves more communication costs than most previous data collection schemes,where they collect n or kn data ***,we conducted extensive experiments to evaluate the UAV-ITD *** results show that compared with previous schemes,our scheme can reduce estimated truth error by 52.25%–96.09%,increase the accuracy of workers’trust evaluation by 0.68–61.82 times,and save recruitment costs by 24.08%–54.15%in truth data discovery.

关键词： Unmanned aerial systems Trust computing Truth discovery Deep matrix factorization Low-cost communications

来源：评论

学校读者我要写书评

暂无评论

Faster AMEDA-A Hybrid Mesoscale Eddy Detection Algorithm

引用

computer Modeling in Engineering & sciences 2024年第11期141卷 1827-1846页

作者： Xinchang Zhang Xiaokang Pan Rongjie Zhu Runda Guan Zhongfeng Qiu Biao Song Organization Department Nanjing University of Information Science and TechnologyNanjing210044China School of Computer Science Nanjing University of Information Science and TechnologyNanjing210044China School of Teacher Education Nanjing University of Information Science and TechnologyNanjing210044China School of Electronics&Information Engineering Nanjing University of Information Science and TechnologyNanjing210044China School of Software Nanjing University of Information Science and TechnologyNanjing210044China

Identification of ocean eddies from a large amount of ocean data provided by satellite measurements and numerical simulations is crucial,while the academia has invented many traditional physical methods with accurate detection capability,but their detection computational efficiency is *** recent years,with the increasing application of deep learning in ocean feature detection,many deep learning-based eddy detection models have been developed for more effective eddy detection from ocean *** it is difficult for them to precisely fit some physical features implicit in traditional methods,leading to inaccurate identification of ocean *** this study,to address the low efficiency of traditional physical methods and the low detection accuracy of deep learning models,we propose a solution that combines the target detection model Faster Region with CNN feature(Faster R-CNN)with the traditional dynamic algorithm Angular Momentum Eddy Detection and Tracking Algorithm(AMEDA).We use Faster R-CNN to detect and generate bounding boxes for eddies,allowing AMEDA to detect the eddy center within these bounding boxes,thus reducing the complexity of center *** demonstrate the detection efficiency and accuracy of this model,this paper compares the experimental results with AMEDA and the deep learning-based eddy detection method *** results show that the eddy detection results of this paper are more accurate than eddyNet and have higher execution efficiency than AMEDA.

关键词： Mesoscale eddy detection object detection contour detection remote sensing

来源：评论

学校读者我要写书评

暂无评论

Domain generalization with semi-supervised learning for people-centric activity recognition

引用

science China(Information sciences) 2025年第1期68卷 171-188页

作者： Jing LIU Wei ZHU Di LI Xing HU Liang SONG Academy for Engineering & Technology Fudan University Shanghai East-bund Research Institute on Networking Systems of AI School of Optoelectronic Information and Computer Engineering University of Shanghai for Science & Technology

People-centric activity recognition is one of the most critical technologies in a wide range of real-world applications,including intelligent transportation systems, healthcare services, and brain-computer interfaces. Large-scale data collection and annotation make the application of machine learning algorithms prohibitively expensive when adapting to new tasks. One way of circumventing this limitation is to train the model in a semi-supervised learning manner that utilizes a percentage of unlabeled data to reduce the labeling burden in prediction tasks. Despite their appeal, these models often assume that labeled and unlabeled data come from similar distributions, which leads to the domain shift problem caused by the presence of distribution gaps. To address these limitations, we propose herein a novel method for people-centric activity recognition,called domain generalization with semi-supervised learning(DGSSL), that effectively enhances the representation learning and domain alignment capabilities of a model. We first design a new autoregressive discriminator for adversarial training between unlabeled and labeled source domains, extracting domain-specific features to reduce the distribution gaps. Second, we introduce two reconstruction tasks to capture the task-specific features to avoid losing information related to representation learning while maintaining task-specific consistency. Finally, benefiting from the collaborative optimization of these two tasks, the model can accurately predict both the domain and category labels of the source domains for the classification task. We conduct extensive experiments on three real-world sensing datasets. The experimental results show that DGSSL surpasses the three state-of-the-art methods with better performance and generalization.

关键词： activity recognition deep learning domain generalization semi-supervised learning adversarial training

来源：评论

学校读者我要写书评

暂无评论

A Lightweight Object Detection Algorithm for Remote Sensing Images

引用

Engineering Letters 2025年第3期33卷 704-711页

作者： Hou, Donghao Zhang, Yujun Ren, Jia School of Computer and Software Engineering University of Science and Technology Liaoning Anshan114051 China School of Electronic and Information Engineering University of Science and Technology Liaoning Anshan114051 China

With the continuous advancement of satellite technology, remote sensing images has been increasingly applied in fields such as urban planning, environmental monitoring, and disaster response. However, remote sensing images often feature small target sizes and complex backgrounds, posing significant computational challenges for object detection tasks. To address this issue, this paper proposes a lightweight remote sensing images object detection algorithm based on YOLOv9. The proposed algorithm incorporates the SimRMB module, which effectively reduces computational complexity while improving the efficiency and accuracy of feature extraction. Through a dynamic attention mechanism, SimRMB is capable of focusing on important regions while minimizing background interference, and by integrating residual learning and skip connections, it ensures the stability of deep networks. To further enhance detection performance, the FasterRepNCSPELAN4 module is introduced, which employs PConv operations to reduce computational load and memory usage. It also utilizes dilated convolutions and DFC attention mechanisms to strengthen feature extraction, thereby increasing the efficiency and accuracy of object detection. Additionally, this study integrates the GhostModuleV2 module, which generates core feature maps and employs lightweight operations to create redundant features, greatly reducing the computational complexity of *** results show that on the SIMD dataset, the improved YOLOv9 model has a parameter size of 167.88 MB and GFLOPs of 208.6. Compared to the baseline YOLOv9 model (parameter size: 194.57 MB, GFLOPs: 239.0), the parameter size is reduced by 13.71%, GFLOPs are reduced by 12.72%, and detection accuracy is improved by 1.4%. These results demonstrate that the proposed lightweight YOLOv9 model effectively reduces computational overhead while maintaining excellent detection performance, providing an efficient solution for object detection tasks in resou

关键词： Satellite imagery

来源：评论

学校读者我要写书评

暂无评论

A Dual-Layer Attention Based CAPTCHA Recognition Approach with Guided Visual Attention

引用

computer Modeling in Engineering & sciences 2025年第3期142卷 2841-2867页

作者： Zaid Derea Beiji Zou Xiaoyan Kui Alaa Thobhani Amr Abdussalam School of Computer Science and Engineering Central South UniversityChangsha410083China College of Computer Science and Information Technology Wasit UniversityWasit52001Iraq Electronic Engineering and Information Science Department University of Science and Technology of ChinaHefei230026China

Enhancing website security is crucial to combat malicious activities,and CAPTCHA(Completely Automated Public Turing tests to tell computers and Humans Apart)has become a key method to distinguish humans from *** text-based CAPTCHAs are designed to challenge machines while remaining human-readable,recent advances in deep learning have enabled models to recognize them with remarkable *** this regard,we propose a novel two-layer visual attention framework for CAPTCHA recognition that builds on traditional attention mechanisms by incorporating Guided Visual Attention(GVA),which sharpens focus on relevant visual *** have specifically adapted the well-established image captioning task to address this *** approach utilizes the first-level attention module as guidance to the second-level attention component,incorporating two LSTM(Long Short-Term Memory)layers to enhance CAPTCHA *** extensive evaluation across four diverse datasets—Weibo,BoC(Bank of China),Gregwar,and Captcha 0.3—shows the adaptability and efficacy of our *** approach demonstrated impressive performance,achieving an accuracy of 96.70%for BoC and 95.92%for *** results underscore the effectiveness of our method in accurately recognizing and processing CAPTCHA datasets,showcasing its robustness,reliability,and ability to handle varied challenges in CAPTCHA recognition.

关键词： Text-based CAPTCHA image recognition guided visual attention web security computer vision

来源：评论

学校读者我要写书评

暂无评论

Augmented reality without SLAM

引用

Multimedia Tools and Applications 2024年 1-17页

作者： Gholami, Aminreza Nasihatkon, Behrooz Soryani, Mohsen School of Computer Engineering Iran University of Science and Technology Tehran Iran School of Computer Engineering K. N. Toosi University of Technology Tehran Iran

Most augmented reality (AR) pipelines typically involve the computation of the camera’s pose in each frame, followed by the 2D projection of virtual objects. The camera pose estimation is commonly implemented as SLAM (Simultaneous Localisation and Mapping) algorithm. However, SLAM systems are often limited to scenarios where the camera intrinsics remain fixed or are known in all frames. This paper presents an initial effort to circumvent the pose estimation stage altogether and directly computes 2D projections using epipolar constraints. To achieve this, we initially calculate the fundamental matrices between the keyframes and each new frame. The 2D locations of objects can then be triangulated by finding the intersection of epipolar lines in the new frame. We propose a robust algorithm that can handle situations where some of the fundamental matrices are entirely erroneous. Most notably, we introduce a depth-buffering algorithm that relies solely on the fundamental matrices, eliminating the need to compute 3D point locations in the target view. By utilizing fundamental matrices, our method remains effective even when all intrinsic camera parameters vary over time. Notably, our proposed approach achieved sufficient accuracy, even with more degrees of freedom in the solution space. © The Author(s), under exclusive licence to Springer science+Business Media, LLC, part of Springer Nature 2024.

关键词： Augmented reality

来源：评论

学校读者我要写书评

暂无评论

Harnessing Trend Theory to Enhance Distributed Proximal Point Algorithm Approaches for Multi-Area Economic Dispatch Optimization

引用

computers, Materials & Continua 2025年第3期82卷 4503-4533页

作者： Yaming Ren Xing Deng College of Mechanical and Control Engineering Guilin University of TechnologyGuilin541006China School of Computer Jiangsu University of Science and TechnologyZhenjiang212003China

The exponential growth in the scale of power systems has led to a significant increase in the complexity of dispatch problem resolution,particularly within multi-area interconnected power *** complexity necessitates the employment of distributed solution methodologies,which are not only essential but also highly *** the realm of computational modelling,the multi-area economic dispatch problem(MAED)can be formulated as a linearly constrained separable convex optimization *** proximal point algorithm(PPA)is particularly adept at addressing such mathematical constructs *** study introduces parallel(PPPA)and serial(SPPA)variants of the PPA as distributed algorithms,specifically designed for the computational modelling of the *** PPA introduces a quadratic term into the objective function,which,while potentially complicating the iterative updates of the algorithm,serves to dampen oscillations near the optimal solution,thereby enhancing the convergence ***,the convergence efficiency of the PPA is significantly influenced by the parameter *** address this parameter sensitivity,this research draws on trend theory from stock market analysis to propose trend theory-driven distributed PPPA and SPPA,thereby enhancing the robustness of the computational *** computational models proposed in this study are anticipated to exhibit superior performance in terms of convergence behaviour,stability,and robustness with respect to parameter selection,potentially outperforming existing methods such as the alternating direction method of multipliers(ADMM)and Auxiliary Problem Principle(APP)in the computational simulation of power system dispatch *** simulation results demonstrate that the trend theory-based PPPA,SPPA,ADMM and APP exhibit significant robustness to the initial value of parameter c,and show superior convergence characteristics compared to the residual balancing ADMM.

关键词： Multi-area economic dispatch problem proximal point algorithm trend theory

来源：评论

学校读者我要写书评

暂无评论

STPM: Spatial-Temporal Token Pruning and Merging for Complex Activity Recognition

引用

IEEE Transactions on Circuits and Systems for Video technology 2024年第6期35卷 5407-5418页

作者： Su, Yumeng Zhang, Jiachao Yan, Rui Li, Pengpeng Xie, Guosen Shu, Xiangbo Nanjing University of Science and Technology School of Computer Science and Engineering Nanjing210094 China Nanjing Institute of Technology Artificial Intelligence Industrial Technology Research Institute Nanjing211167 China Nanjing University Department of Computer Science and Technology Nanjing210023 China

Lightweight video representation techniques have advanced significantly for simple activity recognition, but they still encounter several issues when applied to complex activity recognition: (i) The presence of numerous individuals and varying spatial positions makes it difficult for traditional token pruning methods to maintain accuracy. (ii) Simply discarding entire frames may result in the loss of crucial clues. (iii) To maintain parallel computing, applying the same pruning rate to every frame leads to significant redundancy in frames with low information content. To this end, we propose a lightweight and novel Spatial-Temporal Token Pruning and Merging (STPM) framework, specifically designed for complex action videos where human actors occupy a small spatial resolution within video frames. Our framework considers two critical factors: semantic importance and spatial-temporal redundancy, to further reduce overhead. For semantic importance, STPM captures class-specific attention scores by learning multiple class tokens within the transformer to guide token pruning. For spatial-temporal redundancy, STPM employs an anchor graph and temporal attention to perform spatial and temporal token merging, preserving appearance and temporal cues while eliminating semantic duplication and redundancy. We conduct extensive experiments on JRDB-PAR primarily using recently introduced video transformer backbones, e.g., MViT and ViT. Our framework achieves similar results while requiring 40% less computation. © 1991-2012 IEEE.

关键词： Economic and social effects

来源：评论

学校读者我要写书评

暂无评论

Enhancement of charging performance of quantum battery via quantum coherence of bath

引用

Chinese Physics B 2023年第1期32卷 161-171页

作者：于文莉张允李海魏广芬韩丽萍田峰邹建 School of Computer Science and Technology Shandong Technology and Business UniversityYantai 264005China School of Information and Electronic Engineering Shandong Technology and Business UniversityYantai 264005China School of Science Tianjin University of TechnologyTianjin 300384China School of Physics Beijing Institute of TechnologyBeijing 100081China

An open quantum battery(QB)model of a single qubit system charging in a coherent auxiliary bath(CAB)consisting of a series of independent coherent ancillae is *** to the collision charging protocol we derive a quantum master equation and obtain the analytical solution of QB in a steady *** find that the full charging capacity(or the maximal extractable work(MEW))of QB,in the weak QB-ancilla coupling limit,is positively correlated with the coherence magnitude of *** with the numerical simulations we compare with the charging properties of QB at finite coupling strength,such as the MEW,average charging power and the charging efficiency,when considering the bath to be a thermal auxiliary bath(TAB)and a CAB,*** find that when the QB with CAB,in the weak coupling regime,is in fully charging,both its capacity and charging efficiency can go beyond its classical counterpart,and they increase with the increase of coherence magnitude of *** addition,the MEW of QB in the regime of relative strong coupling and strong coherent magnitude shows the oscillatory behavior with the charging time increasing,and the first peak value can even be larger than the full charging MEW of *** also leads to a much larger average charging power than that of QB with TAB in a short-time charging *** features suggest that with the help of quantum coherence of CAB it becomes feasible to switch the charging schemes between the long-time slow charging protocol with large capacity and high efficiency and the short-time rapid charging protocol with highly charging power only by adjusting the coupling strength of *** work clearly demonstrates that the quantum coherence of bath can not only serve as the role of“fuel”of QB to be utilized to improve the QB's charging performance but also provide an alternative way to integrate the different charging protocols into a single QB.

关键词： quantum battery quantum coherence maximal extractable work charging power

来源：评论

学校读者我要写书评

暂无评论

Lexicon-Augmented Cross-Domain Chinese Word Segmentation with Graph Convolutional Network

引用

Chinese Journal of Electronics 2022年第5期31卷 949-957页

作者： YU Hao HUANG Kaiyu WANG Yu HUANG Degen School of Computer Science and Technology Dalian University of Technology School of Foreign Languages Dalian University of Technology

Existing neural approaches have achieved significant progress for Chinese word segmentation(CWS). The performances of these methods tend to drop dramatically in the cross-domain scenarios due to the data distribution mismatch across domains and the out of vocabulary words problem. To address these two issues,proposes a lexicon-augmented graph convolutional network for cross-domain CWS. The novel model can capture the information of word boundaries from all candidate words and utilize domain lexicons to alleviate the distribution gap across domains. Experimental results on the cross-domain CWS datasets(SIGHAN-2010 and TCM)show that the proposed method successfully models information of domain lexicons for neural CWS approaches and helps to achieve competitive performance for crossdomain CWS. The two problems of cross-domain CWS can be effectively solved through various interactions between characters and candidate words based on ***, experiments on the CWS benchmarks(Bakeoff-2005) also demonstrate the robustness and efficiency of the proposed method.

关键词： cross-domain CWS datasets candidate words neural CWS approaches vocabulary words problem text analysis vocabulary convolutional neural nets domain lexicons natural language processing lexicon-augmented cross-domain Chinese word segmentation graph neural networks lexicon-augmented graph convolutional network learning (artificial intelligence) cross-domain scenarios word processing word boundaries

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：