Underwater pipeline inspection plays a vital role in the proactive maintenance and management of critical marine infrastructure and subaquatic systems.However,the inspection of underwater pipelines presents a challeng...Underwater pipeline inspection plays a vital role in the proactive maintenance and management of critical marine infrastructure and subaquatic systems.However,the inspection of underwater pipelines presents a challenge due to factors such as light scattering,absorption,restricted visibility,and ambient noise.The advancement of deep learning has introduced powerful techniques for processing large amounts of unstructured and imperfect data collected from underwater environments.This study evaluated the efficacy of the You Only Look Once(YOLO)algorithm,a real-time object detection and localization model based on convolutional neural networks,in identifying and classifying various types of pipeline defects in underwater settings.YOLOv8,the latest evolution in the YOLO family,integrates advanced capabilities,such as anchor-free detection,a cross-stage partial network backbone for efficient feature extraction,and a feature pyramid network+path aggregation network neck for robust multi-scale object detection,which make it particularly well-suited for complex underwater environments.Due to the lack of suitable open-access datasets for underwater pipeline defects,a custom dataset was captured using a remotely operated vehicle in a controlled environment.This application has the following assets available for use.Extensive experimentation demonstrated that YOLOv8 X-Large consistently outperformed other models in terms of pipe defect detection and classification and achieved a strong balance between precision and recall in identifying pipeline cracks,rust,corners,defective welds,flanges,tapes,and holes.This research establishes the baseline performance of YOLOv8 for underwater defect detection and showcases its potential to enhance the reliability and efficiency of pipeline inspection tasks in challenging underwater environments.展开更多
As deep learning(DL)models are increasingly deployed in sensitive domains(e.g.,healthcare),concerns over privacy and security have intensified.Conventional penetration testing frameworks,such asOWASP and NIST,are effe...As deep learning(DL)models are increasingly deployed in sensitive domains(e.g.,healthcare),concerns over privacy and security have intensified.Conventional penetration testing frameworks,such asOWASP and NIST,are effective for traditional networks and applications but lack the capabilities to address DL-specific threats,such asmodel inversion,membership inference,and adversarial attacks.This review provides a comprehensive analysis of penetration testing for the privacy of DL models,examining the shortfalls of existing frameworks,tools,and testing methodologies.Through systematic evaluation of existing literature and empirical analysis,we identify three major contributions:(i)a critical assessment of traditional penetration testing frameworks’inadequacies when applied to DL-specific privacy vulnerabilities,(ii)a comprehensive evaluation of state-of-the-art privacy-preserving methods and their integration with penetration testing workflows,and(iii)the development of a structured framework that combines reconnaissance,threat modeling,exploitation,and post-exploitation phases specifically tailored for DL privacy assessment.Moreover,this review evaluates popular solutions such as IBMAdversarial Robustness Toolbox and TensorFlowPrivacy,alongside privacy-preserving techniques(e.g.,Differential Privacy,Homomorphic Encryption,and Federated Learning),which we systematically analyze through comparative studies of their effectiveness,computational overhead,and practical deployment constraints.While these techniques offer promising safeguards,their adoption is hindered by accuracy loss,performance overheads,and the rapid evolution of attack strategies.Our findings reveal that no single existing solution provides comprehensive protection,which leads us to propose a hybrid approach that strategically combines multiple privacy-preserving mechanisms.The findings of this survey underscore an urgent need for automated,regulationcompliant penetration testing frameworks specifically tailored to DL systems.We argue for hybrid privacy solutions that combinemultiple protectivemechanisms to ensure bothmodel accuracy and privacy.Building on our analysis,we present actionable recommendations for developing adaptive penetration testing strategies that incorporate automated vulnerability assessment,continuous monitoring,and regulatory compliance verification.展开更多
As urbanization continues to accelerate,the challenges associated with managing transportation in metropolitan areas become increasingly complex.The surge in population density contributes to traffic congestion,impact...As urbanization continues to accelerate,the challenges associated with managing transportation in metropolitan areas become increasingly complex.The surge in population density contributes to traffic congestion,impacting travel experiences and posing safety risks.Smart urban transportation management emerges as a strategic solution,conceptualized here as a multidimensional big data problem.The success of this strategy hinges on the effective collection of information from diverse,extensive,and heterogeneous data sources,necessitating the implementation of full⁃stack Information and Communication Technology(ICT)solutions.The main idea of the work is to investigate the current technologies of Intelligent Transportation Systems(ITS)and enhance the safety of urban transportation systems.Machine learning models,trained on historical data,can predict traffic congestion,allowing for the implementation of preventive measures.Deep learning architectures,with their ability to handle complex data representations,further refine traffic predictions,contributing to more accurate and dynamic transportation management.The background of this research underscores the challenges posed by traffic congestion in metropolitan areas and emphasizes the need for advanced technological solutions.By integrating GPS and GIS technologies with machine learning algorithms,this work aims to pay attention to the development of intelligent transportation systems that not only address current challenges but also pave the way for future advancements in urban transportation management.展开更多
With the proliferation of Internet of Things(IoT)devices,securing these interconnected systems against cyberattacks has become a critical challenge.Traditional security paradigms often fail to cope with the scale and ...With the proliferation of Internet of Things(IoT)devices,securing these interconnected systems against cyberattacks has become a critical challenge.Traditional security paradigms often fail to cope with the scale and diversity of IoT network traffic.This paper presents a comparative benchmark of classic machine learning(ML)and state-of-the-art deep learning(DL)algorithms for IoT intrusion detection.Our methodology employs a twophased approach:a preliminary pilot study using a custom-generated dataset to establish baselines,followed by a comprehensive evaluation on the large-scale CICIoTDataset2023.We benchmarked algorithms including Random Forest,XGBoost,CNN,and StackedLSTM.The results indicate that while top-performingmodels frombothcategories achieve over 99%classification accuracy,this metric masks a crucial performance trade-off.We demonstrate that treebased ML ensembles exhibit superior precision(91%)in identifying benign traffic,making them effective at reducing false positives.Conversely,DL models demonstrate superior recall(96%),making them better suited for minimizing the interruption of legitimate traffic.We conclude that the selection of an optimal model is not merely a matter of maximizing accuracy but is a strategic choice dependent on the specific security priority either minimizing false alarms or ensuring service availability.Thiswork provides a practical framework for deploying context-aware security solutions in diverse IoT environments.展开更多
Test case prioritization and ranking play a crucial role in software testing by improving fault detection efficiency and ensuring software reliability.While prioritization selects the most relevant test cases for opti...Test case prioritization and ranking play a crucial role in software testing by improving fault detection efficiency and ensuring software reliability.While prioritization selects the most relevant test cases for optimal coverage,ranking further refines their execution order to detect critical faults earlier.This study investigates machine learning techniques to enhance both prioritization and ranking,contributing to more effective and efficient testing processes.We first employ advanced feature engineering alongside ensemble models,including Gradient Boosted,Support Vector Machines,Random Forests,and Naive Bayes classifiers to optimize test case prioritization,achieving an accuracy score of 0.98847 and significantly improving the Average Percentage of Fault Detection(APFD).Subsequently,we introduce a deep Q-learning framework combined with a Genetic Algorithm(GA)to refine test case ranking within priority levels.This approach achieves a rank accuracy of 0.9172,demonstrating robust performance despite the increasing computational demands of specialized variation operators.Our findings highlight the effectiveness of stacked ensemble learning and reinforcement learning in optimizing test case prioritization and ranking.This integrated approach improves testing efficiency,reduces late-stage defects,and improves overall software stability.The study provides valuable information for AI-driven testing frameworks,paving the way for more intelligent and adaptive software quality assurance methodologies.展开更多
Accurate streamflow prediction is essential for flood warning,reservoir operation,irrigation scheduling,hydropower planning,and sustainable water management,yet remains challenging due to the complexity of hydrologica...Accurate streamflow prediction is essential for flood warning,reservoir operation,irrigation scheduling,hydropower planning,and sustainable water management,yet remains challenging due to the complexity of hydrological processes.Although data-driven models often outperform conventional physics-based hydrological modelling approaches,their real-world deployment is limited by cost,infrastructure demands,and the interdisciplinary expertise required.To bridge this gap,this study developed QPred,a regional,lightweight,cost-effective,web-delivered application for daily streamflow forecasting.The study executed an end-to-end workflow,from field data acquisition to accessible web-based deployment for on-demand forecasting.High-resolution rainfall data were recorded with tippingbucket gauges and loggers,while river water depth in the Aglar and Paligaad watersheds was converted to discharge using site-specific rating curves,resulting in a daily dataset of precipitation,river water level and discharge.Four DL architectures were trained,including vanilla Long Short-Term Memory(LSTM),stacked LSTM,bidirectional LSTM,and Gated Recurrent Unit(GRU),and evaluated using Nash-Sutcliffe Efficiency(NSE),Coefficient of Determination(R2),Root-Mean-Square-Error-Standard-Deviation Ratio(RSR),and Percentage Bias(PBIAS)metrics.Performance was watershed-specific,as the vanilla LSTM demonstrated the best generalisation for the Aglar watershed(R2=0.88,NSE=0.82,RMSE=0.12 during validation),while the GRU achieved the highest validation accuracy in Paligaad(R2=0.88,NSE=0.88,RMSE=0.49).All models achieved satisfactory to excellent performance during calibration(R2>0.91,NSE>0.91 for both watersheds),demonstrating strong capability to capture streamflow dynamics.The highest performing models were selected and embedded into the QPred application.QPred was developed as a lightweight web pipeline,utilising Google Colab as the primary execution environment,Flask as the backend inference framework,Google Drive for artefact storage,andNgrok for secureHTTPS tunnelling.Auser-friendly front end utilises range sliders(bounded by observed minima and maxima)to gather inputs and provides discharge data along with metadata,thereby enhancing transparency.This work demonstrates that accurate,context-aware deep learningmodels can be delivered through low-cost,web-based platforms,providing a reproducible and scalable pipeline for hydrological applications in other watersheds and for practitioners.展开更多
Accurate detection of driver fatigue is essential for improving road safety.This study investigates the effectiveness of using multimodal physiological signals for fatigue detection while incorporating uncertainty qua...Accurate detection of driver fatigue is essential for improving road safety.This study investigates the effectiveness of using multimodal physiological signals for fatigue detection while incorporating uncertainty quantification to enhance the reliability of predictions.Physiological signals,including Electrocardiogram(ECG),Galvanic Skin Response(GSR),and Electroencephalogram(EEG),were transformed into image representations and analyzed using pretrained deep neu-ral networks.The extracted features were classified through a feedforward neural network,and prediction reliability was assessed using uncertainty quantification techniques such as Monte Carlo Dropout(MCD),model ensembles,and combined approaches.Evaluation metrics included standard measures(sensitivity,specificity,precision,and accuracy)along with uncertainty-aware metrics such as uncertainty sensitivity and uncertainty precision.Across all evaluations,ECG-based models consistently demonstrated strong performance.The findings indicate that combining multimodal physi-ological signals,Transfer Learning(TL),and uncertainty quantification can significantly improve both the accuracy and trustworthiness of fatigue detection systems.This approach supports the development of more reliable driver assistance technologies aimed at preventing fatigue-related accidents.展开更多
Nondestructive measurement technology of phenotype can provide substantial phenotypic data support for applications such as seedling breeding,management,and quality testing.The current method of measuring seedling phe...Nondestructive measurement technology of phenotype can provide substantial phenotypic data support for applications such as seedling breeding,management,and quality testing.The current method of measuring seedling phenotypes mainly relies on manual measurement which is inefficient,subjective and destroys samples.Therefore,the paper proposes a nondestructive measurement method for the canopy phenotype of the watermelon plug seedlings based on deep learning.The Azure Kinect was used to shoot canopy color images,depth images,and RGB-D images of the watermelon plug seedlings.The Mask-RCNN network was used to classify,segment,and count the canopy leaves of the watermelon plug seedlings.To reduce the error of leaf area measurement caused by mutual occlusion of leaves,the leaves were repaired by CycleGAN,and the depth images were restored by image processing.Then,the Delaunay triangulation was adopted to measure the leaf area in the leaf point cloud.The YOLOX target detection network was used to identify the growing point position of each seedling on the plug tray.Then the depth differences between the growing point and the upper surface of the plug tray were calculated to obtain plant height.The experiment results show that the nondestructive measurement algorithm proposed in this paper achieves good measurement performance for the watermelon plug seedlings from the 1 true-leaf to 3 true-leaf stages.The average relative error of measurement is 2.33%for the number of true leaves,4.59%for the number of cotyledons,8.37%for the leaf area,and 3.27%for the plant height.The experiment results demonstrate that the proposed algorithm in this paper provides an effective solution for the nondestructive measurement of the canopy phenotype of the plug seedlings.展开更多
Intelligent inspection of transmission lines enables efficient automated fault detection by integrating artificial intelligence,robotics,and other related technologies.It plays a key role in ensuring power grid safety...Intelligent inspection of transmission lines enables efficient automated fault detection by integrating artificial intelligence,robotics,and other related technologies.It plays a key role in ensuring power grid safety,reducing operation and maintenance costs,driving the digital transformation of the power industry,and facilitating the achievement of the dual-carbon goals.This review focuses on vision-based power line inspection,with deep learning as the core perspective to systematically analyze the latest research advancements in this field.Firstly,at the technical foundation level,it elaborates on deep learning algorithms for intelligent transmission line inspection based on image perception,covering object detection algorithms,semantic segmentation algorithms,and other relevant methodologies.Secondly,in application practice,it summarizes deep learning-based intelligent inspection applications across six dimensions—including detection of power insulators and their defects,transmission tower detection,power line feature extraction,metal fitting and defect detection,thermal fault diagnosis of power components,and safety hazard detection in power scenarios,and further lists relevant public datasets.Finally,in response to current challenges,it identifies five key future research directions,such as the deep integration of multiple learning paradigms,multi-modal data fusion,collaborative application of large and small models,cloud-edge-end collaborative integration,and multi-agent cluster control.This paper reviews and analyzes numerous deep learning-based intelligent detectionmethods for aerial images,comprehensively explores the application of deep learning in Unmanned Aerial Vehicle(UAV)inspection scenarios,and thus provides valuable theoretical and practical references for scholars engaged in smart grid automated inspection research.展开更多
Deep learning attentionmechanisms have achieved remarkable progress in computer vision,but still face limitations when handling images with ambiguous boundaries and uncertain feature representations.Conventional atten...Deep learning attentionmechanisms have achieved remarkable progress in computer vision,but still face limitations when handling images with ambiguous boundaries and uncertain feature representations.Conventional attention modules such as SE-Net,CBAM,ECA-Net,and CA adopt a deterministic paradigm,assigning fixed scalar weights to features without modeling ambiguity or confidence.To overcome these limitations,this paper proposes the Fuzzy Attention Network Layer(FANL),which integrates intuitionistic fuzzy set theory with convolutional neural networks to explicitly represent feature uncertainty through membership(μ),non-membership(ν),and hesitation(π)degrees.FANLconsists of four coremodules:(1)feature dimensionality reduction via global pooling,(2)fuzzymodeling using learnable clustering centers,(3)adaptive attention generation through weighted fusion of fuzzy components,and(4)feature refinement through residual connections.A cross-layer guidance mechanism is further introduced to enhance hierarchical feature propagation,allowing high-level semantic features to incorporate fine-grained texture information from shallow layers.Comprehensive experiments on three benchmark datasets—PathMNIST-30000,full PathMNIST,and Blood MNIST—demonstrate the effectiveness and generalizability of FANL.The model achieves 84.41±0.56%accuracy and a 1.69%improvement over the baseline CNN while maintaining lightweight computational complexity.Ablation studies show that removing any component causes a 1.7%–2.0%performance drop,validating the synergistic contribution of each module.Furthermore,FANL provides superior uncertainty calibration(ECE=0.0452)and interpretable selective prediction under uncertainty.Overall,FANL presents an efficient and uncertaintyaware attention framework that improves both accuracy and reliability,offering a promising direction for robust visual recognition under ambiguous or noisy conditions.展开更多
Breast cancer screening programs rely heavily on mammography for early detection;however,diagnostic performance is strongly affected by inter-reader variability,breast density,and the limitations of conven-tional comp...Breast cancer screening programs rely heavily on mammography for early detection;however,diagnostic performance is strongly affected by inter-reader variability,breast density,and the limitations of conven-tional computer-aided detection systems.Recent advances in deep learning have enabled more robust and scalable solutions for large-scale screening,yet a systematic comparison of modern object detection architectures on nationally representative datasets remains limited.This study presents a comprehensive quantitative comparison of prominent deep learning–based object detection architectures for Artificial Intelligence-assisted mammography analysis using the MammosighTR dataset,developed within the Turkish National Breast Cancer Screening Program.The dataset comprises 12,740 patient cases collected between 2016 and 2022,annotated with BI-RADS categories,breast density levels,and lesion localization labels.A total of 31 models were evaluated,including One-Stage,Two-Stage,and Transformer-based architectures,under a unified experimental framework at both patient and breast levels.The results demonstrate that Two-Stage architectures consistently outperform One-Stage models,achieving approximately 2%–4%higher Macro F1-Scores and more balanced precision–recall trade-offs,with Double-Head R-CNN and Dynamic R-CNN yielding the highest overall performance(Macro F1≈0.84–0.86).This advantage is primarily attributed to the region proposal mechanism and improved class balance inherent to Two-Stage designs.One-Stage detectors exhibited higher sensitivity and faster inference,reaching Recall values above 0.88,but experienced minor reductions in Precision and overall accuracy(≈1%–2%)compared with Two-Stage models.Among Transformer-based architectures,Deformable DEtection TRansformer demonstrated strong robustness and consistency across datasets,achieving Macro F1-Scores comparable to CNN-based detectors(≈0.83–0.85)while exhibiting minimal performance degradation under distributional shifts.Breast density–based analysis revealed increased misclassification rates in medium-density categories(types B and C),whereas Transformer-based architectures maintained more stable performance in high-density type D tissue.These findings quantitatively confirm that both architectural design and tissue characteristics play a decisive role in diagnostic accuracy.Overall,the study provides a reproducible benchmark and highlights the potential of hybrid approaches that combine the accuracy of Two-Stage detectors with the contextual modeling capability of Transformer architectures for clinically reliable breast cancer screening systems.展开更多
In the context of rural revitalization and the development of smart agriculture, image classification technology based on deep learning has emerged as a crucial tool for digital monitoring and intelligent prevention a...In the context of rural revitalization and the development of smart agriculture, image classification technology based on deep learning has emerged as a crucial tool for digital monitoring and intelligent prevention and control of agricultural diseases. This paper provides a systematic review of the evolutionary development of algorithms within this field. Addressing challenges such as domain drift and limited global awareness in classical convolutional neural networks (CNNs) applied to complex agricultural environments, the paper focuses on the latest advancements in vision transformers (ViT) and their hybrid architectures to enhance cross-domain robustness and fine-grained recognition capabilities. In response to the challenges posed by scarce long-tail data and limited edge computing power in real-world scenarios, the paper explores solutions related to few-shot learning and ultra-lightweight network deployment. Finally, a forward-looking analysis is presented on the application paradigms of multimodal feature fusion, vision-based large models, and explainable artificial intelligence (AI) within smart plant protection. This analysis aims to offer theoretical insights for the development of efficient and transparent intelligent diagnostic systems for agricultural diseases, thereby supporting the advancement of digital agriculture and the construction of a robust agricultural nation.展开更多
Waves are important physical phenomena in an ocean,and their accurate prediction is essential for ocean engineering,maritime traffic,and marine early warning systems.This study focuses on the Qinhuangdao Sea area loca...Waves are important physical phenomena in an ocean,and their accurate prediction is essential for ocean engineering,maritime traffic,and marine early warning systems.This study focuses on the Qinhuangdao Sea area located in the Bohai Sea,China.Herein,we use on-site wind data to correct the reanalysis wind data obtained from the European Centre for Medium-Range Weather Forecasts(ECMWF),improving the accuracy of boundary conditions.Then,we use the Simulating WAves Nearshore(SWAN)model to simulate the regional wave field over time.A regional wave-parameter prediction model is then developed using a limited number of sampled data(covering only 2 years,2020–2021);the model is based on the Whale Optimization Algorithm(WOA),convolutional neural networks(CNNs),and long short-term memory(LSTM)neural networks.WOA is used to optimize the CNN and LSTM framework;in this framework,CNN extracts spatial features,and the LSTM network captures temporal features,enabling accurate short and long-term predictions of wave height,period,and direction.The experimental results showed that despite the small sample size,the model achieves a goodness of fit of 0.9957 for wave height prediction,0.9973 for period,and 0.9749 for wave direction in short-term forecasting.As the prediction step size increases,the accuracy of the model decreases.When the prediction step size reaches 9 h,the root mean square error for the prediction of wave height,period,and direction increases to 0.2060 m,0.4582 s,and32.5358°,respectively.The reliability and applicability of the model are further validated by the experimental results.Our findings highlighted the potential of the developed model in operational wave forecasting,even with a limited number of sampled data.展开更多
Human Activity Recognition(HAR)is a novel area for computer vision.It has a great impact on healthcare,smart environments,and surveillance while is able to automatically detect human behavior.It plays a vital role in ...Human Activity Recognition(HAR)is a novel area for computer vision.It has a great impact on healthcare,smart environments,and surveillance while is able to automatically detect human behavior.It plays a vital role in many applications,such as smart home,healthcare,human computer interaction,sports analysis,and especially,intelligent surveillance.In this paper,we propose a robust and efficient HAR system by leveraging deep learning paradigms,including pre-trained models,CNN architectures,and their average-weighted fusion.However,due to the diversity of human actions and various environmental influences,as well as a lack of data and resources,achieving high recognition accuracy remain elusive.In this work,a weighted average ensemble technique is employed to fuse three deep learning models:EfficientNet,ResNet50,and a custom CNN.The results of this study indicate that using a weighted average ensemble strategy for developing more effective HAR models may be a promising idea for detection and classification of human activities.Experiments by using the benchmark dataset proved that the proposed weighted ensemble approach outperformed existing approaches in terms of accuracy and other key performance measures.The combined average-weighted ensemble of pre-trained and CNN models obtained an accuracy of 98%,compared to 97%,96%,and 95%for the customized CNN,EfficientNet,and ResNet50 models,respectively.展开更多
Structural Health Monitoring(SHM)plays a critical role in ensuring the safety,integrity,longevity and economic efficiency of civil infrastructures.The field has undergone a profound transformation over the last few de...Structural Health Monitoring(SHM)plays a critical role in ensuring the safety,integrity,longevity and economic efficiency of civil infrastructures.The field has undergone a profound transformation over the last few decades,evolving from traditional methods—often reliant on visual inspections—to data-driven intelligent systems.This review paper analyzes this historical trajectory,beginning with the approaches that relied on modal parameters as primary damage indicators.The advent of advanced sensor technologies and increased computational power brings a significant change,making Machine Learning(ML)a viable and powerful tool for damage assessment.More recently,Deep Learning(DL)has emerged as a paradigm shift,allowing for more automated processing of large data sets(such as the structural vibration signals and other types of sensors)with excellent performance and accuracy,often surpassing previous methods.This paper systematically reviews these technological milestones—from traditional vibration-based methods to the current state-of-the-art in deep learning.Finally,it critically examines emerging trends—such as Digital Twins and Transformer-based architectures—and discusses future research directions that will shape the next generation of SHM systems for civil engineering.展开更多
To address the high costs and operational instability of distribution networks caused by the large-scale integration of distributed energy resources(DERs)(such as photovoltaic(PV)systems,wind turbines(WT),and energy s...To address the high costs and operational instability of distribution networks caused by the large-scale integration of distributed energy resources(DERs)(such as photovoltaic(PV)systems,wind turbines(WT),and energy storage(ES)devices),and the increased grid load fluctuations and safety risks due to uncoordinated electric vehicles(EVs)charging,this paper proposes a novel dual-scale hierarchical collaborative optimization strategy.This strategy decouples system-level economic dispatch from distributed EV agent control,effectively solving the resource coordination conflicts arising from the high computational complexity,poor scalability of existing centralized optimization,or the reliance on local information decision-making in fully decentralized frameworks.At the lower level,an EV charging and discharging model with a hybrid discrete-continuous action space is established,and optimized using an improved Parameterized Deep Q-Network(PDQN)algorithm,which directly handles mode selection and power regulation while embedding physical constraints to ensure safety.At the upper level,microgrid(MG)operators adopt a dynamic pricing strategy optimized through Deep Reinforcement Learning(DRL)to maximize economic benefits and achieve peak-valley shaving.Simulation results show that the proposed strategy outperforms traditional methods,reducing the total operating cost of the MG by 21.6%,decreasing the peak-to-valley load difference by 33.7%,reducing the number of voltage limit violations by 88.9%,and lowering the average electricity cost for EV users by 15.2%.This method brings a win-win result for operators and users,providing a reliable and efficient scheduling solution for distribution networks with high renewable energy penetration rates.展开更多
The exponential growth of video content has driven significant advancements in video summarization techniques in recent years. Breakthroughs in deep learning have been particularly transformative, enabling more effect...The exponential growth of video content has driven significant advancements in video summarization techniques in recent years. Breakthroughs in deep learning have been particularly transformative, enabling more effective detection of key information and creating new possibilities for video synopsis. To summarize recent progress and accelerate research in this field,this paper provides a comprehensive review of deep learningbased video summarization methods developed over the past decade. We begin by examining the research landscape of video abstraction technologies and identifying core challenges in video summarization. Subsequently, we systematically analyze prevailing deep learning frameworks and methodologies employed in current video summarization systems, offering researchers a clear roadmap of the field's evolution. Unlike previous review works,we first classify research papers based on the structural hierarchy of the video(from frame-level to shot-level to video-level),then further categorize them according to the summary backbone model(feature extraction and spatiotemporal modeling).This approach provides a more systematic and hierarchical organization of the documents. Following this comprehensive review,we summarize the benchmark datasets and evaluation metrics commonly employed in the field. Finally, we analyze persistent challenges and propose insightful directions for future research,providing a forward-looking perspective on video summarization technologies. This systematic literature review is of great reference value to new researchers exploring the fields of deep learning and video summarization.展开更多
Intrusion detection in Internet of Things(IoT)environments presents challenges due to heterogeneous devices,diverse attack vectors,and highly imbalanced datasets.Existing research on the ToN-IoT dataset has largely em...Intrusion detection in Internet of Things(IoT)environments presents challenges due to heterogeneous devices,diverse attack vectors,and highly imbalanced datasets.Existing research on the ToN-IoT dataset has largely emphasized binary classification and single-model pipelines,which often showstrong performance but limited generalizability,probabilistic reliability,and operational interpretability.This study proposes a stacked ensemble deep learning framework that integrates random forest,extreme gradient boosting,and a deep neural network as base learners,with CatBoost as the meta-learner.On the ToN-IoT Linux process dataset,the model achieved near-perfect discrimination(macro area under the curve=0.998),robust calibration,and superior F1-scores compared with standalone classifiers.Interpretability was achieved through SHapley Additive exPlanations–based feature attribution,which highlights actionable drivers ofmalicious behavior,such as command-line patterns,process scheduling anomalies,and CPU usage spikes,and aligns these indicators with MITRE ATT&CK tactics and techniques.Complementary analyses,including cumulative lift and sensitivity-specificity trade-offs,revealed the framework’s suitability for deployment in security operations centers,where calibrated risk scores,transparent explanations,and resource-aware triage are essential.These contributions bridge methodological rigor in artificial intelligence/machine learning with operational priorities in cybersecurity,delivering a scalable and explainable intrusion detection system suitable for real-world deployment in IoT environments.展开更多
Effective partitioning is crucial for enabling parallel restoration of power systems after blackouts.This paper proposes a novel partitioning method based on deep reinforcement learning.First,the partitioning decision...Effective partitioning is crucial for enabling parallel restoration of power systems after blackouts.This paper proposes a novel partitioning method based on deep reinforcement learning.First,the partitioning decision process is formulated as a Markov decision process(MDP)model to maximize the modularity.Corresponding key partitioning constraints on parallel restoration are considered.Second,based on the partitioning objective and constraints,the reward function of the partitioning MDP model is set by adopting a relative deviation normalization scheme to reduce mutual interference between the reward and penalty in the reward function.The soft bonus scaling mechanism is introduced to mitigate overestimation caused by abrupt jumps in the reward.Then,the deep Q network method is applied to solve the partitioning MDP model and generate partitioning schemes.Two experience replay buffers are employed to speed up the training process of the method.Finally,case studies on the IEEE 39-bus test system demonstrate that the proposed method can generate a high-modularity partitioning result that meets all key partitioning constraints,thereby improving the parallelism and reliability of the restoration process.Moreover,simulation results demonstrate that an appropriate discount factor is crucial for ensuring both the convergence speed and the stability of the partitioning training.展开更多
The annual compliance cycle of the carbon trading system allows generation companies(GenCos)to decouple the timing of carbon allowance purchases from their actual emissions.However,trading a large volume of allowances...The annual compliance cycle of the carbon trading system allows generation companies(GenCos)to decouple the timing of carbon allowance purchases from their actual emissions.However,trading a large volume of allowances within a single day can significantly impact on carbon prices.Faced with uncertain future carbon and electricity prices,GenCos must address a challenging multistage stochastic optimization problem to coordinate their carbon trading strategies with daily power generation decisions.In this paper,a two-layered hybrid mathematical-deep reinforcement learning(DRL)optimization framework is proposed.The upper DRL layer tackles the stochastic,year-long carbon trading and allowance usage optimization problem,aiming for long-term optimality and providing guidance for short-term decisions in the lower layer.The lower mathematical optimization layer addresses the deterministic daily power generation schedule problem while enforcing strict technical constraints.To accelerate learning of the annual compliance cycle,a decision timeline transfer learning method is proposed,enabling the DRL agent to progressively refine its policy through sequentially training on monthly,weekly and daily decision environments.Case studies demonstrate that,with these methods,a GenCo can reduce emission costs and increase profits by effectively leveraging carbon price fluctuations within the compliance cycle.展开更多
文摘Underwater pipeline inspection plays a vital role in the proactive maintenance and management of critical marine infrastructure and subaquatic systems.However,the inspection of underwater pipelines presents a challenge due to factors such as light scattering,absorption,restricted visibility,and ambient noise.The advancement of deep learning has introduced powerful techniques for processing large amounts of unstructured and imperfect data collected from underwater environments.This study evaluated the efficacy of the You Only Look Once(YOLO)algorithm,a real-time object detection and localization model based on convolutional neural networks,in identifying and classifying various types of pipeline defects in underwater settings.YOLOv8,the latest evolution in the YOLO family,integrates advanced capabilities,such as anchor-free detection,a cross-stage partial network backbone for efficient feature extraction,and a feature pyramid network+path aggregation network neck for robust multi-scale object detection,which make it particularly well-suited for complex underwater environments.Due to the lack of suitable open-access datasets for underwater pipeline defects,a custom dataset was captured using a remotely operated vehicle in a controlled environment.This application has the following assets available for use.Extensive experimentation demonstrated that YOLOv8 X-Large consistently outperformed other models in terms of pipe defect detection and classification and achieved a strong balance between precision and recall in identifying pipeline cracks,rust,corners,defective welds,flanges,tapes,and holes.This research establishes the baseline performance of YOLOv8 for underwater defect detection and showcases its potential to enhance the reliability and efficiency of pipeline inspection tasks in challenging underwater environments.
基金supported in part by the Tianjin Natural Science Foundation Project(24JCZDJC01000)the Fundamental Research Funds for the Central Universities of China(No.3122025091).
文摘As deep learning(DL)models are increasingly deployed in sensitive domains(e.g.,healthcare),concerns over privacy and security have intensified.Conventional penetration testing frameworks,such asOWASP and NIST,are effective for traditional networks and applications but lack the capabilities to address DL-specific threats,such asmodel inversion,membership inference,and adversarial attacks.This review provides a comprehensive analysis of penetration testing for the privacy of DL models,examining the shortfalls of existing frameworks,tools,and testing methodologies.Through systematic evaluation of existing literature and empirical analysis,we identify three major contributions:(i)a critical assessment of traditional penetration testing frameworks’inadequacies when applied to DL-specific privacy vulnerabilities,(ii)a comprehensive evaluation of state-of-the-art privacy-preserving methods and their integration with penetration testing workflows,and(iii)the development of a structured framework that combines reconnaissance,threat modeling,exploitation,and post-exploitation phases specifically tailored for DL privacy assessment.Moreover,this review evaluates popular solutions such as IBMAdversarial Robustness Toolbox and TensorFlowPrivacy,alongside privacy-preserving techniques(e.g.,Differential Privacy,Homomorphic Encryption,and Federated Learning),which we systematically analyze through comparative studies of their effectiveness,computational overhead,and practical deployment constraints.While these techniques offer promising safeguards,their adoption is hindered by accuracy loss,performance overheads,and the rapid evolution of attack strategies.Our findings reveal that no single existing solution provides comprehensive protection,which leads us to propose a hybrid approach that strategically combines multiple privacy-preserving mechanisms.The findings of this survey underscore an urgent need for automated,regulationcompliant penetration testing frameworks specifically tailored to DL systems.We argue for hybrid privacy solutions that combinemultiple protectivemechanisms to ensure bothmodel accuracy and privacy.Building on our analysis,we present actionable recommendations for developing adaptive penetration testing strategies that incorporate automated vulnerability assessment,continuous monitoring,and regulatory compliance verification.
文摘As urbanization continues to accelerate,the challenges associated with managing transportation in metropolitan areas become increasingly complex.The surge in population density contributes to traffic congestion,impacting travel experiences and posing safety risks.Smart urban transportation management emerges as a strategic solution,conceptualized here as a multidimensional big data problem.The success of this strategy hinges on the effective collection of information from diverse,extensive,and heterogeneous data sources,necessitating the implementation of full⁃stack Information and Communication Technology(ICT)solutions.The main idea of the work is to investigate the current technologies of Intelligent Transportation Systems(ITS)and enhance the safety of urban transportation systems.Machine learning models,trained on historical data,can predict traffic congestion,allowing for the implementation of preventive measures.Deep learning architectures,with their ability to handle complex data representations,further refine traffic predictions,contributing to more accurate and dynamic transportation management.The background of this research underscores the challenges posed by traffic congestion in metropolitan areas and emphasizes the need for advanced technological solutions.By integrating GPS and GIS technologies with machine learning algorithms,this work aims to pay attention to the development of intelligent transportation systems that not only address current challenges but also pave the way for future advancements in urban transportation management.
文摘With the proliferation of Internet of Things(IoT)devices,securing these interconnected systems against cyberattacks has become a critical challenge.Traditional security paradigms often fail to cope with the scale and diversity of IoT network traffic.This paper presents a comparative benchmark of classic machine learning(ML)and state-of-the-art deep learning(DL)algorithms for IoT intrusion detection.Our methodology employs a twophased approach:a preliminary pilot study using a custom-generated dataset to establish baselines,followed by a comprehensive evaluation on the large-scale CICIoTDataset2023.We benchmarked algorithms including Random Forest,XGBoost,CNN,and StackedLSTM.The results indicate that while top-performingmodels frombothcategories achieve over 99%classification accuracy,this metric masks a crucial performance trade-off.We demonstrate that treebased ML ensembles exhibit superior precision(91%)in identifying benign traffic,making them effective at reducing false positives.Conversely,DL models demonstrate superior recall(96%),making them better suited for minimizing the interruption of legitimate traffic.We conclude that the selection of an optimal model is not merely a matter of maximizing accuracy but is a strategic choice dependent on the specific security priority either minimizing false alarms or ensuring service availability.Thiswork provides a practical framework for deploying context-aware security solutions in diverse IoT environments.
文摘Test case prioritization and ranking play a crucial role in software testing by improving fault detection efficiency and ensuring software reliability.While prioritization selects the most relevant test cases for optimal coverage,ranking further refines their execution order to detect critical faults earlier.This study investigates machine learning techniques to enhance both prioritization and ranking,contributing to more effective and efficient testing processes.We first employ advanced feature engineering alongside ensemble models,including Gradient Boosted,Support Vector Machines,Random Forests,and Naive Bayes classifiers to optimize test case prioritization,achieving an accuracy score of 0.98847 and significantly improving the Average Percentage of Fault Detection(APFD).Subsequently,we introduce a deep Q-learning framework combined with a Genetic Algorithm(GA)to refine test case ranking within priority levels.This approach achieves a rank accuracy of 0.9172,demonstrating robust performance despite the increasing computational demands of specialized variation operators.Our findings highlight the effectiveness of stacked ensemble learning and reinforcement learning in optimizing test case prioritization and ranking.This integrated approach improves testing efficiency,reduces late-stage defects,and improves overall software stability.The study provides valuable information for AI-driven testing frameworks,paving the way for more intelligent and adaptive software quality assurance methodologies.
文摘Accurate streamflow prediction is essential for flood warning,reservoir operation,irrigation scheduling,hydropower planning,and sustainable water management,yet remains challenging due to the complexity of hydrological processes.Although data-driven models often outperform conventional physics-based hydrological modelling approaches,their real-world deployment is limited by cost,infrastructure demands,and the interdisciplinary expertise required.To bridge this gap,this study developed QPred,a regional,lightweight,cost-effective,web-delivered application for daily streamflow forecasting.The study executed an end-to-end workflow,from field data acquisition to accessible web-based deployment for on-demand forecasting.High-resolution rainfall data were recorded with tippingbucket gauges and loggers,while river water depth in the Aglar and Paligaad watersheds was converted to discharge using site-specific rating curves,resulting in a daily dataset of precipitation,river water level and discharge.Four DL architectures were trained,including vanilla Long Short-Term Memory(LSTM),stacked LSTM,bidirectional LSTM,and Gated Recurrent Unit(GRU),and evaluated using Nash-Sutcliffe Efficiency(NSE),Coefficient of Determination(R2),Root-Mean-Square-Error-Standard-Deviation Ratio(RSR),and Percentage Bias(PBIAS)metrics.Performance was watershed-specific,as the vanilla LSTM demonstrated the best generalisation for the Aglar watershed(R2=0.88,NSE=0.82,RMSE=0.12 during validation),while the GRU achieved the highest validation accuracy in Paligaad(R2=0.88,NSE=0.88,RMSE=0.49).All models achieved satisfactory to excellent performance during calibration(R2>0.91,NSE>0.91 for both watersheds),demonstrating strong capability to capture streamflow dynamics.The highest performing models were selected and embedded into the QPred application.QPred was developed as a lightweight web pipeline,utilising Google Colab as the primary execution environment,Flask as the backend inference framework,Google Drive for artefact storage,andNgrok for secureHTTPS tunnelling.Auser-friendly front end utilises range sliders(bounded by observed minima and maxima)to gather inputs and provides discharge data along with metadata,thereby enhancing transparency.This work demonstrates that accurate,context-aware deep learningmodels can be delivered through low-cost,web-based platforms,providing a reproducible and scalable pipeline for hydrological applications in other watersheds and for practitioners.
基金the Australian Research Council Discovery Projects funding scheme(DP190102181,DP210101465).
文摘Accurate detection of driver fatigue is essential for improving road safety.This study investigates the effectiveness of using multimodal physiological signals for fatigue detection while incorporating uncertainty quantification to enhance the reliability of predictions.Physiological signals,including Electrocardiogram(ECG),Galvanic Skin Response(GSR),and Electroencephalogram(EEG),were transformed into image representations and analyzed using pretrained deep neu-ral networks.The extracted features were classified through a feedforward neural network,and prediction reliability was assessed using uncertainty quantification techniques such as Monte Carlo Dropout(MCD),model ensembles,and combined approaches.Evaluation metrics included standard measures(sensitivity,specificity,precision,and accuracy)along with uncertainty-aware metrics such as uncertainty sensitivity and uncertainty precision.Across all evaluations,ECG-based models consistently demonstrated strong performance.The findings indicate that combining multimodal physi-ological signals,Transfer Learning(TL),and uncertainty quantification can significantly improve both the accuracy and trustworthiness of fatigue detection systems.This approach supports the development of more reliable driver assistance technologies aimed at preventing fatigue-related accidents.
基金funded by the National Key Research and Development Program of China(Grant No.2019YFD1001900)the HZAU-AGIS Cooperation Fund(Grant No.SZYJY2022006).
文摘Nondestructive measurement technology of phenotype can provide substantial phenotypic data support for applications such as seedling breeding,management,and quality testing.The current method of measuring seedling phenotypes mainly relies on manual measurement which is inefficient,subjective and destroys samples.Therefore,the paper proposes a nondestructive measurement method for the canopy phenotype of the watermelon plug seedlings based on deep learning.The Azure Kinect was used to shoot canopy color images,depth images,and RGB-D images of the watermelon plug seedlings.The Mask-RCNN network was used to classify,segment,and count the canopy leaves of the watermelon plug seedlings.To reduce the error of leaf area measurement caused by mutual occlusion of leaves,the leaves were repaired by CycleGAN,and the depth images were restored by image processing.Then,the Delaunay triangulation was adopted to measure the leaf area in the leaf point cloud.The YOLOX target detection network was used to identify the growing point position of each seedling on the plug tray.Then the depth differences between the growing point and the upper surface of the plug tray were calculated to obtain plant height.The experiment results show that the nondestructive measurement algorithm proposed in this paper achieves good measurement performance for the watermelon plug seedlings from the 1 true-leaf to 3 true-leaf stages.The average relative error of measurement is 2.33%for the number of true leaves,4.59%for the number of cotyledons,8.37%for the leaf area,and 3.27%for the plant height.The experiment results demonstrate that the proposed algorithm in this paper provides an effective solution for the nondestructive measurement of the canopy phenotype of the plug seedlings.
基金financially supported by theNatural Research Project of College in Anhui Province under grant 2024AH051365,2025AHGXZK30826Research Platform of New Energy and Energy-Saving Technology Research Center under grant KYJG002.
文摘Intelligent inspection of transmission lines enables efficient automated fault detection by integrating artificial intelligence,robotics,and other related technologies.It plays a key role in ensuring power grid safety,reducing operation and maintenance costs,driving the digital transformation of the power industry,and facilitating the achievement of the dual-carbon goals.This review focuses on vision-based power line inspection,with deep learning as the core perspective to systematically analyze the latest research advancements in this field.Firstly,at the technical foundation level,it elaborates on deep learning algorithms for intelligent transmission line inspection based on image perception,covering object detection algorithms,semantic segmentation algorithms,and other relevant methodologies.Secondly,in application practice,it summarizes deep learning-based intelligent inspection applications across six dimensions—including detection of power insulators and their defects,transmission tower detection,power line feature extraction,metal fitting and defect detection,thermal fault diagnosis of power components,and safety hazard detection in power scenarios,and further lists relevant public datasets.Finally,in response to current challenges,it identifies five key future research directions,such as the deep integration of multiple learning paradigms,multi-modal data fusion,collaborative application of large and small models,cloud-edge-end collaborative integration,and multi-agent cluster control.This paper reviews and analyzes numerous deep learning-based intelligent detectionmethods for aerial images,comprehensively explores the application of deep learning in Unmanned Aerial Vehicle(UAV)inspection scenarios,and thus provides valuable theoretical and practical references for scholars engaged in smart grid automated inspection research.
文摘Deep learning attentionmechanisms have achieved remarkable progress in computer vision,but still face limitations when handling images with ambiguous boundaries and uncertain feature representations.Conventional attention modules such as SE-Net,CBAM,ECA-Net,and CA adopt a deterministic paradigm,assigning fixed scalar weights to features without modeling ambiguity or confidence.To overcome these limitations,this paper proposes the Fuzzy Attention Network Layer(FANL),which integrates intuitionistic fuzzy set theory with convolutional neural networks to explicitly represent feature uncertainty through membership(μ),non-membership(ν),and hesitation(π)degrees.FANLconsists of four coremodules:(1)feature dimensionality reduction via global pooling,(2)fuzzymodeling using learnable clustering centers,(3)adaptive attention generation through weighted fusion of fuzzy components,and(4)feature refinement through residual connections.A cross-layer guidance mechanism is further introduced to enhance hierarchical feature propagation,allowing high-level semantic features to incorporate fine-grained texture information from shallow layers.Comprehensive experiments on three benchmark datasets—PathMNIST-30000,full PathMNIST,and Blood MNIST—demonstrate the effectiveness and generalizability of FANL.The model achieves 84.41±0.56%accuracy and a 1.69%improvement over the baseline CNN while maintaining lightweight computational complexity.Ablation studies show that removing any component causes a 1.7%–2.0%performance drop,validating the synergistic contribution of each module.Furthermore,FANL provides superior uncertainty calibration(ECE=0.0452)and interpretable selective prediction under uncertainty.Overall,FANL presents an efficient and uncertaintyaware attention framework that improves both accuracy and reliability,offering a promising direction for robust visual recognition under ambiguous or noisy conditions.
文摘Breast cancer screening programs rely heavily on mammography for early detection;however,diagnostic performance is strongly affected by inter-reader variability,breast density,and the limitations of conven-tional computer-aided detection systems.Recent advances in deep learning have enabled more robust and scalable solutions for large-scale screening,yet a systematic comparison of modern object detection architectures on nationally representative datasets remains limited.This study presents a comprehensive quantitative comparison of prominent deep learning–based object detection architectures for Artificial Intelligence-assisted mammography analysis using the MammosighTR dataset,developed within the Turkish National Breast Cancer Screening Program.The dataset comprises 12,740 patient cases collected between 2016 and 2022,annotated with BI-RADS categories,breast density levels,and lesion localization labels.A total of 31 models were evaluated,including One-Stage,Two-Stage,and Transformer-based architectures,under a unified experimental framework at both patient and breast levels.The results demonstrate that Two-Stage architectures consistently outperform One-Stage models,achieving approximately 2%–4%higher Macro F1-Scores and more balanced precision–recall trade-offs,with Double-Head R-CNN and Dynamic R-CNN yielding the highest overall performance(Macro F1≈0.84–0.86).This advantage is primarily attributed to the region proposal mechanism and improved class balance inherent to Two-Stage designs.One-Stage detectors exhibited higher sensitivity and faster inference,reaching Recall values above 0.88,but experienced minor reductions in Precision and overall accuracy(≈1%–2%)compared with Two-Stage models.Among Transformer-based architectures,Deformable DEtection TRansformer demonstrated strong robustness and consistency across datasets,achieving Macro F1-Scores comparable to CNN-based detectors(≈0.83–0.85)while exhibiting minimal performance degradation under distributional shifts.Breast density–based analysis revealed increased misclassification rates in medium-density categories(types B and C),whereas Transformer-based architectures maintained more stable performance in high-density type D tissue.These findings quantitatively confirm that both architectural design and tissue characteristics play a decisive role in diagnostic accuracy.Overall,the study provides a reproducible benchmark and highlights the potential of hybrid approaches that combine the accuracy of Two-Stage detectors with the contextual modeling capability of Transformer architectures for clinically reliable breast cancer screening systems.
基金Supported by School-level Project of Shaoyang Industry Polytechnic College(SKY24A06)Science and Technology Plan(Special Fund Subsidy)of Shaoyang City(2024PT4070)General Research Project of Hunan Provincial Department of Education in 2025(25C1457).
文摘In the context of rural revitalization and the development of smart agriculture, image classification technology based on deep learning has emerged as a crucial tool for digital monitoring and intelligent prevention and control of agricultural diseases. This paper provides a systematic review of the evolutionary development of algorithms within this field. Addressing challenges such as domain drift and limited global awareness in classical convolutional neural networks (CNNs) applied to complex agricultural environments, the paper focuses on the latest advancements in vision transformers (ViT) and their hybrid architectures to enhance cross-domain robustness and fine-grained recognition capabilities. In response to the challenges posed by scarce long-tail data and limited edge computing power in real-world scenarios, the paper explores solutions related to few-shot learning and ultra-lightweight network deployment. Finally, a forward-looking analysis is presented on the application paradigms of multimodal feature fusion, vision-based large models, and explainable artificial intelligence (AI) within smart plant protection. This analysis aims to offer theoretical insights for the development of efficient and transparent intelligent diagnostic systems for agricultural diseases, thereby supporting the advancement of digital agriculture and the construction of a robust agricultural nation.
基金supported by the National Natural Science Foundation of China(Nos.52071057,52171247)the Liaoning Youth Elite Talent Program(No.XLYC220309)。
文摘Waves are important physical phenomena in an ocean,and their accurate prediction is essential for ocean engineering,maritime traffic,and marine early warning systems.This study focuses on the Qinhuangdao Sea area located in the Bohai Sea,China.Herein,we use on-site wind data to correct the reanalysis wind data obtained from the European Centre for Medium-Range Weather Forecasts(ECMWF),improving the accuracy of boundary conditions.Then,we use the Simulating WAves Nearshore(SWAN)model to simulate the regional wave field over time.A regional wave-parameter prediction model is then developed using a limited number of sampled data(covering only 2 years,2020–2021);the model is based on the Whale Optimization Algorithm(WOA),convolutional neural networks(CNNs),and long short-term memory(LSTM)neural networks.WOA is used to optimize the CNN and LSTM framework;in this framework,CNN extracts spatial features,and the LSTM network captures temporal features,enabling accurate short and long-term predictions of wave height,period,and direction.The experimental results showed that despite the small sample size,the model achieves a goodness of fit of 0.9957 for wave height prediction,0.9973 for period,and 0.9749 for wave direction in short-term forecasting.As the prediction step size increases,the accuracy of the model decreases.When the prediction step size reaches 9 h,the root mean square error for the prediction of wave height,period,and direction increases to 0.2060 m,0.4582 s,and32.5358°,respectively.The reliability and applicability of the model are further validated by the experimental results.Our findings highlighted the potential of the developed model in operational wave forecasting,even with a limited number of sampled data.
基金supported by Princess Nourah bint Abdulrahman University Researchers Supporting Project number(PNURSP2026R765),Princess Nourah bint Abdulrahman University,Riyadh,Saudi Arabia.
文摘Human Activity Recognition(HAR)is a novel area for computer vision.It has a great impact on healthcare,smart environments,and surveillance while is able to automatically detect human behavior.It plays a vital role in many applications,such as smart home,healthcare,human computer interaction,sports analysis,and especially,intelligent surveillance.In this paper,we propose a robust and efficient HAR system by leveraging deep learning paradigms,including pre-trained models,CNN architectures,and their average-weighted fusion.However,due to the diversity of human actions and various environmental influences,as well as a lack of data and resources,achieving high recognition accuracy remain elusive.In this work,a weighted average ensemble technique is employed to fuse three deep learning models:EfficientNet,ResNet50,and a custom CNN.The results of this study indicate that using a weighted average ensemble strategy for developing more effective HAR models may be a promising idea for detection and classification of human activities.Experiments by using the benchmark dataset proved that the proposed weighted ensemble approach outperformed existing approaches in terms of accuracy and other key performance measures.The combined average-weighted ensemble of pre-trained and CNN models obtained an accuracy of 98%,compared to 97%,96%,and 95%for the customized CNN,EfficientNet,and ResNet50 models,respectively.
基金The authors would like to thank CNPq(Conselho Nacional de Desenvolvimento Científico e Tecnológico)—grants 407256/2022-9,303550/2025-2,402533/2023-2 and 303982/2022-5FAPEMIG(Fundação de AmparoàPesquisa do Estado de Minas Gerais)—grants APQ-00032-24 and APD-01113-25 for their financial support.
文摘Structural Health Monitoring(SHM)plays a critical role in ensuring the safety,integrity,longevity and economic efficiency of civil infrastructures.The field has undergone a profound transformation over the last few decades,evolving from traditional methods—often reliant on visual inspections—to data-driven intelligent systems.This review paper analyzes this historical trajectory,beginning with the approaches that relied on modal parameters as primary damage indicators.The advent of advanced sensor technologies and increased computational power brings a significant change,making Machine Learning(ML)a viable and powerful tool for damage assessment.More recently,Deep Learning(DL)has emerged as a paradigm shift,allowing for more automated processing of large data sets(such as the structural vibration signals and other types of sensors)with excellent performance and accuracy,often surpassing previous methods.This paper systematically reviews these technological milestones—from traditional vibration-based methods to the current state-of-the-art in deep learning.Finally,it critically examines emerging trends—such as Digital Twins and Transformer-based architectures—and discusses future research directions that will shape the next generation of SHM systems for civil engineering.
基金supported in part by the Research on Key Technologies for the Development of an Active Balancing Cooperative Control Systemfor Distribution Networks and the National Natural Science Foundation of China under Grant 521532240029,Grant 62303006.
文摘To address the high costs and operational instability of distribution networks caused by the large-scale integration of distributed energy resources(DERs)(such as photovoltaic(PV)systems,wind turbines(WT),and energy storage(ES)devices),and the increased grid load fluctuations and safety risks due to uncoordinated electric vehicles(EVs)charging,this paper proposes a novel dual-scale hierarchical collaborative optimization strategy.This strategy decouples system-level economic dispatch from distributed EV agent control,effectively solving the resource coordination conflicts arising from the high computational complexity,poor scalability of existing centralized optimization,or the reliance on local information decision-making in fully decentralized frameworks.At the lower level,an EV charging and discharging model with a hybrid discrete-continuous action space is established,and optimized using an improved Parameterized Deep Q-Network(PDQN)algorithm,which directly handles mode selection and power regulation while embedding physical constraints to ensure safety.At the upper level,microgrid(MG)operators adopt a dynamic pricing strategy optimized through Deep Reinforcement Learning(DRL)to maximize economic benefits and achieve peak-valley shaving.Simulation results show that the proposed strategy outperforms traditional methods,reducing the total operating cost of the MG by 21.6%,decreasing the peak-to-valley load difference by 33.7%,reducing the number of voltage limit violations by 88.9%,and lowering the average electricity cost for EV users by 15.2%.This method brings a win-win result for operators and users,providing a reliable and efficient scheduling solution for distribution networks with high renewable energy penetration rates.
基金supported by UKRI(EP/Z000025/1)Horizon Europe Programme under the MSCA grant for the ACMod project(101130271)。
文摘The exponential growth of video content has driven significant advancements in video summarization techniques in recent years. Breakthroughs in deep learning have been particularly transformative, enabling more effective detection of key information and creating new possibilities for video synopsis. To summarize recent progress and accelerate research in this field,this paper provides a comprehensive review of deep learningbased video summarization methods developed over the past decade. We begin by examining the research landscape of video abstraction technologies and identifying core challenges in video summarization. Subsequently, we systematically analyze prevailing deep learning frameworks and methodologies employed in current video summarization systems, offering researchers a clear roadmap of the field's evolution. Unlike previous review works,we first classify research papers based on the structural hierarchy of the video(from frame-level to shot-level to video-level),then further categorize them according to the summary backbone model(feature extraction and spatiotemporal modeling).This approach provides a more systematic and hierarchical organization of the documents. Following this comprehensive review,we summarize the benchmark datasets and evaluation metrics commonly employed in the field. Finally, we analyze persistent challenges and propose insightful directions for future research,providing a forward-looking perspective on video summarization technologies. This systematic literature review is of great reference value to new researchers exploring the fields of deep learning and video summarization.
文摘Intrusion detection in Internet of Things(IoT)environments presents challenges due to heterogeneous devices,diverse attack vectors,and highly imbalanced datasets.Existing research on the ToN-IoT dataset has largely emphasized binary classification and single-model pipelines,which often showstrong performance but limited generalizability,probabilistic reliability,and operational interpretability.This study proposes a stacked ensemble deep learning framework that integrates random forest,extreme gradient boosting,and a deep neural network as base learners,with CatBoost as the meta-learner.On the ToN-IoT Linux process dataset,the model achieved near-perfect discrimination(macro area under the curve=0.998),robust calibration,and superior F1-scores compared with standalone classifiers.Interpretability was achieved through SHapley Additive exPlanations–based feature attribution,which highlights actionable drivers ofmalicious behavior,such as command-line patterns,process scheduling anomalies,and CPU usage spikes,and aligns these indicators with MITRE ATT&CK tactics and techniques.Complementary analyses,including cumulative lift and sensitivity-specificity trade-offs,revealed the framework’s suitability for deployment in security operations centers,where calibrated risk scores,transparent explanations,and resource-aware triage are essential.These contributions bridge methodological rigor in artificial intelligence/machine learning with operational priorities in cybersecurity,delivering a scalable and explainable intrusion detection system suitable for real-world deployment in IoT environments.
基金funded by the Beijing Engineering Research Center of Electric Rail Transportation.
文摘Effective partitioning is crucial for enabling parallel restoration of power systems after blackouts.This paper proposes a novel partitioning method based on deep reinforcement learning.First,the partitioning decision process is formulated as a Markov decision process(MDP)model to maximize the modularity.Corresponding key partitioning constraints on parallel restoration are considered.Second,based on the partitioning objective and constraints,the reward function of the partitioning MDP model is set by adopting a relative deviation normalization scheme to reduce mutual interference between the reward and penalty in the reward function.The soft bonus scaling mechanism is introduced to mitigate overestimation caused by abrupt jumps in the reward.Then,the deep Q network method is applied to solve the partitioning MDP model and generate partitioning schemes.Two experience replay buffers are employed to speed up the training process of the method.Finally,case studies on the IEEE 39-bus test system demonstrate that the proposed method can generate a high-modularity partitioning result that meets all key partitioning constraints,thereby improving the parallelism and reliability of the restoration process.Moreover,simulation results demonstrate that an appropriate discount factor is crucial for ensuring both the convergence speed and the stability of the partitioning training.
基金supported by the Natural Science Foundation of China-Smart Grid Joint Fund of State Grid Corporation of China(No.U2066212)the Na-tional Natural Science Foundation of China(No.52207105)the Key Science and Technology Pro-jects of China Southern Power Grid Corporation(No.066600KK52222023).
文摘The annual compliance cycle of the carbon trading system allows generation companies(GenCos)to decouple the timing of carbon allowance purchases from their actual emissions.However,trading a large volume of allowances within a single day can significantly impact on carbon prices.Faced with uncertain future carbon and electricity prices,GenCos must address a challenging multistage stochastic optimization problem to coordinate their carbon trading strategies with daily power generation decisions.In this paper,a two-layered hybrid mathematical-deep reinforcement learning(DRL)optimization framework is proposed.The upper DRL layer tackles the stochastic,year-long carbon trading and allowance usage optimization problem,aiming for long-term optimality and providing guidance for short-term decisions in the lower layer.The lower mathematical optimization layer addresses the deterministic daily power generation schedule problem while enforcing strict technical constraints.To accelerate learning of the annual compliance cycle,a decision timeline transfer learning method is proposed,enabling the DRL agent to progressively refine its policy through sequentially training on monthly,weekly and daily decision environments.Case studies demonstrate that,with these methods,a GenCo can reduce emission costs and increase profits by effectively leveraging carbon price fluctuations within the compliance cycle.
