The rapid proliferation of Internet of Things(IoT)devices in critical healthcare infrastructure has introduced significant security and privacy challenges that demand innovative,distributed architectural solutions.Thi...The rapid proliferation of Internet of Things(IoT)devices in critical healthcare infrastructure has introduced significant security and privacy challenges that demand innovative,distributed architectural solutions.This paper proposes FE-ACS(Fog-Edge Adaptive Cybersecurity System),a novel hierarchical security framework that intelligently distributes AI-powered anomaly detection algorithms across edge,fog,and cloud layers to optimize security efficacy,latency,and privacy.Our comprehensive evaluation demonstrates that FE-ACS achieves superior detection performance with an AUC-ROC of 0.985 and an F1-score of 0.923,while maintaining significantly lower end-to-end latency(18.7 ms)compared to cloud-centric(152.3 ms)and fog-only(34.5 ms)architectures.The system exhibits exceptional scalability,supporting up to 38,000 devices with logarithmic performance degradation—a 67×improvement over conventional cloud-based approaches.By incorporating differential privacy mechanisms with balanced privacy-utility tradeoffs(ε=1.0–1.5),FE-ACS maintains 90%–93%detection accuracy while ensuring strong privacy guarantees for sensitive healthcare data.Computational efficiency analysis reveals that our architecture achieves a detection rate of 12,400 events per second with only 12.3 mJ energy consumption per inference.In healthcare risk assessment,FE-ACS demonstrates robust operational viability with low patient safety risk(14.7%)and high system reliability(94.0%).The proposed framework represents a significant advancement in distributed security architectures,offering a scalable,privacy-preserving,and real-time solution for protecting healthcare IoT ecosystems against evolving cyber threats.展开更多
Online examinations have become a dominant assessment mode,increasing concerns over academic integrity.To address the critical challenge of detecting cheating behaviours,this study proposes a hybrid deep learning appr...Online examinations have become a dominant assessment mode,increasing concerns over academic integrity.To address the critical challenge of detecting cheating behaviours,this study proposes a hybrid deep learning approach that combines visual detection and temporal behaviour classification.The methodology utilises object detection models—You Only Look Once(YOLOv12),Faster Region-based Convolutional Neural Network(RCNN),and Single Shot Detector(SSD)MobileNet—integrated with classification models such as Convolutional Neural Networks(CNN),Bidirectional Gated Recurrent Unit(Bi-GRU),and CNN-LSTM(Long Short-Term Memory).Two distinct datasets were used:the Online Exam Proctoring(EOP)dataset from Michigan State University and the School of Computer Science,Duy Tan Unievrsity(SCS-DTU)dataset collected in a controlled classroom setting.A diverse set of cheating behaviours,including book usage,unauthorised interaction,internet access,and mobile phone use,was categorised.Comprehensive experiments evaluated the models based on accuracy,precision,recall,training time,inference speed,and memory usage.We evaluate nine detector-classifier pairings under a unified budget and score them via a calibrated harmonic mean of detection and classification accuracies,enabling deployment-oriented selection under latency and memory constraints.Macro-Precision/Recall/F1 and Receiver Operating Characteristic-Area Under the Curve(ROC-AUC)are reported for the top configurations,revealing consistent advantages of object-centric pipelines for fine-grained cheating cues.The highest overall score is achieved by YOLOv12+CNN(97.15%accuracy),while SSD-MobileNet+CNN provides the best speed-efficiency trade-off for edge devices.This research provides valuable insights into selecting and deploying appropriate deep learning models for maintaining exam integrity under varying resource constraints.展开更多
In fire rescue scenarios,traditional manual operations are highly dangerous,as dense smoke,low visibility,extreme heat,and toxic gases not only hinder rescue efficiency but also endanger firefighters’safety.Although ...In fire rescue scenarios,traditional manual operations are highly dangerous,as dense smoke,low visibility,extreme heat,and toxic gases not only hinder rescue efficiency but also endanger firefighters’safety.Although intelligent rescue robots can enter hazardous environments in place of humans,smoke poses major challenges for human detection algorithms.These challenges include the attenuation of visible and infrared signals,complex thermal fields,and interference frombackground objects,all ofwhichmake it difficult to accurately identify trapped individuals.To address this problem,we propose VIF-YOLO,a visible–infrared fusion model for real-time human detection in dense smoke environments.The framework introduces a lightweight multimodal fusion(LMF)module based on learnable low-rank representation blocks to end-to-end integrate visible and infrared images,preserving fine details while enhancing salient features.In addition,an efficient multiscale attention(EMA)mechanism is incorporated into the YOLOv10n backbone to improve feature representation under low-light conditions.Extensive experiments on our newly constructedmultimodal smoke human detection(MSHD)dataset demonstrate thatVIF-YOLOachievesmAP50 of 99.5%,precision of 99.2%,and recall of 99.3%,outperforming YOLOv10n by a clear margin.Furthermore,when deployed on the NVIDIA Jetson Xavier NX,VIF-YOLO attains 40.6 FPS with an average inference latency of 24.6 ms,validating its real-time capability on edge-computing platforms.These results confirm that VIF-YOLO provides accurate,robust,and fast detection across complex backgrounds and diverse smoke conditions,ensuring reliable and rapid localization of individuals in need of rescue.展开更多
Re-entry gliding vehicles exhibit high maneuverability,making trajectory prediction a key factor in the effectiveness of defense systems.To overcome the limited fitting accuracy of existing methods and their poor adap...Re-entry gliding vehicles exhibit high maneuverability,making trajectory prediction a key factor in the effectiveness of defense systems.To overcome the limited fitting accuracy of existing methods and their poor adaptability to maneuver mode mutations,a trajectory prediction method is proposed that integrates online maneuver mode identification with dynamic modeling.Characteristic parameters are extracted from tracking data for parameterized modeling,enabling real-time identification of maneuver modes.In addition,a maneuver detection mechanism based on higher-order cumulants is introduced to detect lateral maneuver mutations and optimize the use of historical data.Simulation results show that the proposed method achieves accurate trajectory prediction during the glide phase and maintains high accuracy under maneuver mutations,significantly enhancing the prediction performance of both three-dimensional trajectories and ground tracks.展开更多
The Intrusion Detection System(IDS)is a security mechanism developed to observe network traffic and recognize suspicious or malicious activities.Clustering algorithms are often incorporated into IDS;however,convention...The Intrusion Detection System(IDS)is a security mechanism developed to observe network traffic and recognize suspicious or malicious activities.Clustering algorithms are often incorporated into IDS;however,conventional clustering-based methods face notable drawbacks,including poor scalability in handling high-dimensional datasets and a strong dependence of outcomes on initial conditions.To overcome the performance limitations of existing methods,this study proposes a novel quantum-inspired clustering algorithm that relies on a similarity coefficient-based quantum genetic algorithm(SC-QGA)and an improved quantum artificial bee colony algorithm hybrid K-means(IQABC-K).First,the SC-QGA algorithmis constructed based on quantum computing and integrates similarity coefficient theory to strengthen genetic diversity and feature extraction capabilities.For the subsequent clustering phase,the process based on the IQABC-K algorithm is enhanced with the core improvement of adaptive rotation gate and movement exploitation strategies to balance the exploration capabilities of global search and the exploitation capabilities of local search.Simultaneously,the acceleration of convergence toward the global optimum and a reduction in computational complexity are facilitated by means of the global optimum bootstrap strategy and a linear population reduction strategy.Through experimental evaluation with multiple algorithms and diverse performance metrics,the proposed algorithm confirms reliable accuracy on three datasets:KDD CUP99,NSL_KDD,and UNSW_NB15,achieving accuracy of 98.57%,98.81%,and 98.32%,respectively.These results affirm its potential as an effective solution for practical clustering applications.展开更多
The problem of fake news detection(FND)is becoming increasingly important in the field of natural language processing(NLP)because of the rapid dissemination of misleading information on the web.Large language models(L...The problem of fake news detection(FND)is becoming increasingly important in the field of natural language processing(NLP)because of the rapid dissemination of misleading information on the web.Large language models(LLMs)such as GPT-4.Zero excels in natural language understanding tasks but can still struggle to distinguish between fact and fiction,particularly when applied in the wild.However,a key challenge of existing FND methods is that they only consider unimodal data(e.g.,images),while more detailed multimodal data(e.g.,user behaviour,temporal dynamics)is neglected,and the latter is crucial for full-context understanding.To overcome these limitations,we introduce M3-FND(Multimodal Misinformation Mitigation for False News Detection),a novel methodological framework that integrates LLMs with multimodal data sources to perform context-aware veracity assessments.Our method proposes a hybrid system that combines image-text alignment,user credibility profiling,and temporal pattern recognition,which is also strengthened through a natural feedback loop that provides real-time feedback for correcting downstream errors.We use contextual reinforcement learning to schedule prompt updating and update the classifier threshold based on the latest multimodal input,which enables the model to better adapt to changing misinformation attack strategies.M3-FND is tested on three diverse datasets,FakeNewsNet,Twitter15,andWeibo,which contain both text and visual socialmedia content.Experiments showthatM3-FND significantly outperforms conventional and LLMbased baselines in terms of accuracy,F1-score,and AUC on all benchmarks.Our results indicate the importance of employing multimodal cues and adaptive learning for effective and timely detection of fake news.展开更多
A high-sensitivity,low-noise single photon avalanche diode(SPAD)detector was presented based on a 180 nm BCD process.The proposed device utilizes a p-implant layer/high-voltage n-well(HVNW)junction to form a deep aval...A high-sensitivity,low-noise single photon avalanche diode(SPAD)detector was presented based on a 180 nm BCD process.The proposed device utilizes a p-implant layer/high-voltage n-well(HVNW)junction to form a deep avalanche multiplication region for near-infrared(NIR)sensitivity enhancement.By optimizing the device size and electric field of the guard ring,the fill factor(FF)is significantly improved,further increasing photon detection efficiency(PDE).To solve the dark noise caused by the increasing active diameter,a field polysilicon gate structure connected to the p+anode was investigated,effectively suppressing dark count noise by 76.6%.It is experimentally shown that when the active diameter increases from 5 to 10μm,the FF is significantly improved from 20.7%to 39.1%,and thus the peak PDE also rises from 13.3%to 25.8%.At an excess bias voltage of 5 V,a NIR photon detection probability(PDP)of 6.8%at 905 nm,a dark count rate(DCR)of 2.12 cps/μm^(2),an afterpulsing probability(AP)of 1.2%,and a timing jitter of 216 ps are achieved,demonstrating excellent single photon detection performance.展开更多
Infrared(IR)spectroscopy,a technique within the realm of molecular vibrational spectroscopy,furnishes distinctive chemical signatures pivotal for both structural analysis and compound identification.A notable challeng...Infrared(IR)spectroscopy,a technique within the realm of molecular vibrational spectroscopy,furnishes distinctive chemical signatures pivotal for both structural analysis and compound identification.A notable challenge emerges from the misalignment between the mid-IR light wavelength range and molecular dimensions,culminating in a constrained absorption cross-section and diminished vibrational absorption coefficients(Supplementary data).展开更多
Objective:To explore the clinical correlation between the detection of irregular antibodies in red blood cell blood groups and hemolytic disease of the newborn.Methods:This study selected newborns who underwent examin...Objective:To explore the clinical correlation between the detection of irregular antibodies in red blood cell blood groups and hemolytic disease of the newborn.Methods:This study selected newborns who underwent examinations and were diagnosed with hemolytic disease at our hospital from October 2024 to October 2025 as the research subjects.Based on the severity of their hemolytic disease,the infants were divided into a severe group and a mild group.All the infants underwent detection for irregular antibodies in their red blood cell blood groups.General information,blood types,and irregular antibody test results of the two groups were recorded.Univariate analysis was conducted,and variables with statistical significance from the univariate analysis were included in a multivariate logistic regression analysis to explore the clinical correlation between the detection of irregular antibodies in red blood cell blood groups and hemolytic disease of the newborn.Results:Through univariate analysis,it was found that IgG1 and IgG3 subclass antibodies,as well as ABO blood group incompatibility,were statistically significant(p<0.05).When these factors were included in a multivariate logistic regression analysis,it was discovered that IgG1(OR=2.461,95%CI:1.859-2.709),IgG3(OR=2.509,95%CI:1.918-2.893),and ABO blood group incompatibility(OR=2.998,95%CI:2.149-3.493)all exhibited a positive correlation with hemolytic disease of the newborn.Conclusion:As levels of IgG1,IgG3,and ABO blood group incompatibility increase,the incidence of hemolytic disease of the newborn also rises,warranting clinical attention.展开更多
It is of great importance to obtain precise trace data,as traces are frequently the sole visible and measurable parameter in most outcrops.The manual recognition and detection of traces on high-resolution three-dimens...It is of great importance to obtain precise trace data,as traces are frequently the sole visible and measurable parameter in most outcrops.The manual recognition and detection of traces on high-resolution three-dimensional(3D)models are relatively straightforward but time-consuming.One potential solution to enhance this process is to use machine learning algorithms to detect the 3D traces.In this study,a unique pixel-wise texture mapper algorithm generates a dense point cloud representation of an outcrop with the precise resolution of the original textured 3D model.A virtual digital image rendering was then employed to capture virtual images of selected regions.This technique helps to overcome limitations caused by the surface morphology of the rock mass,such as restricted access,lighting conditions,and shading effects.After AI-powered trace detection on two-dimensional(2D)images,a 3D data structuring technique was applied to the selected trace pixels.In the 3D data structuring,the trace data were structured through 2D thinning,3D reprojection,clustering,segmentation,and segment linking.Finally,the linked segments were exported as 3D polylines,with each polyline in the output corresponding to a trace.The efficacy of the proposed method was assessed using a 3D model of a real-world case study,which was used to compare the results of artificial intelligence(AI)-aided and human intelligence trace detection.Rosette diagrams,which visualize the distribution of trace orientations,confirmed the high similarity between the automatically and manually generated trace maps.In conclusion,the proposed semi-automatic method was easy to use,fast,and accurate in detecting the dominant jointing system of the rock mass.展开更多
This paper presents a high-speed and robust dual-band infrared thermal camera based on an ARM CPU.The system consists of a low-resolution long-wavelength infrared detector,a digital temperature and humid⁃ity sensor,an...This paper presents a high-speed and robust dual-band infrared thermal camera based on an ARM CPU.The system consists of a low-resolution long-wavelength infrared detector,a digital temperature and humid⁃ity sensor,and a CMOS sensor.In view of the significant contrast between face and background in thermal infra⁃red images,this paper explores a suitable accuracy-latency tradeoff for thermal face detection and proposes a tiny,lightweight detector named YOLO-Fastest-IR.Four YOLO-Fastest-IR models(IR0 to IR3)with different scales are designed based on YOLO-Fastest.To train and evaluate these lightweight models,a multi-user low-resolution thermal face database(RGBT-MLTF)was collected,and the four networks were trained.Experiments demon⁃strate that the lightweight convolutional neural network performs well in thermal infrared face detection tasks.The proposed algorithm outperforms existing face detection methods in both positioning accuracy and speed,making it more suitable for deployment on mobile platforms or embedded devices.After obtaining the region of interest(ROI)in the infrared(IR)image,the RGB camera is guided by the thermal infrared face detection results to achieve fine positioning of the RGB face.Experimental results show that YOLO-Fastest-IR achieves a frame rate of 92.9 FPS on a Raspberry Pi 4B and successfully detects 97.4%of faces in the RGBT-MLTF test set.Ultimate⁃ly,an infrared temperature measurement system with low cost,strong robustness,and high real-time perfor⁃mance was integrated,achieving a temperature measurement accuracy of 0.3℃.展开更多
Visible and near-infrared photodetectors are widely used in intelligent driving,health monitoring,and other fields.However,the application of photodetectors in the near-infrared region is significantly impacted by hig...Visible and near-infrared photodetectors are widely used in intelligent driving,health monitoring,and other fields.However,the application of photodetectors in the near-infrared region is significantly impacted by high dark current,which can greatly reduce their performance and sensitivity,thereby limiting their effectiveness in certain applications.In this work,the introduction of a C60 back interface layer successfully mitigated back interface reactions to decrease the thickness of the Mo(S,Se)_(2)layer,tailoring the back-contact barrier and preventing reverse charge injection,resulting in a kesterite photodetector with an ultralow dark current density of 5.2×10^(-9)mA/cm^(2)and ultra-weak-light detection at levels as low as 25 pW/cm^(2).Besides,under a self-powered operation,it demonstrates outstanding performance,achieving a peak responsivity of 0.68 A/W,a wide response range spanning from 300 to 1600 nm,and an impressive detectivity of 5.27×10^(14)Jones.In addition,it offers exceptionally rapid response times,with rise and decay times of 70 and 650 ns,respectively.This research offers important insights for developing high-performance self-powered near-infrared photodetectors that have high responsivity,rapid response times,and ultralow dark current.展开更多
As modern power systems grow in complexity,accurate and efficient fault detection has become increasingly important.While many existing reviews focus on a single modality,this paper presents a comprehensive survey fro...As modern power systems grow in complexity,accurate and efficient fault detection has become increasingly important.While many existing reviews focus on a single modality,this paper presents a comprehensive survey from a dual-modality perspective-infrared imaging and voiceprint analysis-two complementary,non-contact techniques that capture different fault characteristics.Infrared imaging excels at detecting thermal anomalies,while voiceprint signals provide insight into mechanical vibrations and internal discharge phenomena.We review both traditional signal processing and deep learning-based approaches for each modality,categorized by key processing stages such as feature extraction and classification.The paper highlights how these modalities address distinct fault types and how they may be fused to improve robustness and accuracy.Representative datasets are summarized,and practical challenges such as noise interference,limited fault samples,and deployment constraints are discussed.By offering a cross-modal,comparative analysis,this work aims to bridge fragmented research and guide future development in intelligent fault detection systems.The review concludes with research trends including multimodal fusion,lightweight models,and self-supervised learning.展开更多
In the context of target detection under infrared conditions for drones,the common issues of high missed detection rates,low signal-to-noise ratio,and blurred edge features for small targets are prevalent.To address t...In the context of target detection under infrared conditions for drones,the common issues of high missed detection rates,low signal-to-noise ratio,and blurred edge features for small targets are prevalent.To address these challenges,this paper proposes an improved detection algorithm based on YOLOv11n.First,a Dynamic Multi-Scale Feature Fusion and Adaptive Weighting approach is employed to design an Adaptive Focused Diffusion Pyramid Network(AFDPN),which enhances the feature expression and transmission capability of shallow small targets,thereby reducing the loss of detailed information.Then,combined with an Edge Enhancement(EE)module,the model improves the extraction of infrared small target edge features through low-frequency suppression and high-frequency enhancement strategies.Experimental results on the publicly available HIT-UAV dataset show that the improved model achieves a 3.8%increase in average detection accuracy and a 3.0%improvement in recall rate compared to YOLOv11n,with a computational cost of only 9.1 GFLOPS.In comparison experiments,the detection accuracy and model size balance achieved the optimal solution,meeting the lightweight deployment requirements for drone-based systems.This method provides a high-precision,lightweight solution for small target detection in drone-based infrared imagery.展开更多
The Global Positioning System(GPS)plays an indispensable role in the control of Unmanned Aerial Vehicle(UAV).However,the civilian GPS signals,transmitted over the air without any encryption,are vulnerable to spoofing ...The Global Positioning System(GPS)plays an indispensable role in the control of Unmanned Aerial Vehicle(UAV).However,the civilian GPS signals,transmitted over the air without any encryption,are vulnerable to spoofing attacks,which further guides the UAV on deviated positions or trajectories.To counter the GPS,,m spoofing on UAV system and to detect the position/trajectory anomaly in real time,a motion state vector based stack long short-term memory trajectory prediction scheme is firstly proposed,leveraging the temporal and spatial features of UAV kinematics.Based on the predicted results,an ensemble voting-based trajectory anomaly detection scheme is proposed to detect the position anomalies in real time with the information of motion state sequences.The proposed prediction-based trajectory anomaly detection scheme outperforms the existing offline detection schemes designed for fixed trajectories.Software In The Loop(SITL)based online prediction and online anomaly detection are demonstrated with random 3D flight trajectories.Results show that the coefficient of determination(R^(2))and Root Mean Square Error(RMSE)of the prediction scheme can reach 0.996 and 3.467,respectively.The accuracy,recall,and F1-score of the proposed anomaly detection scheme can reach 0.984,0.988,and 0.983,respectively,which outperform deep ensemble learning,LSTM-based classifier,machine learning classifier and GA-XGBoost based schemes.Moreover,results show that compared with LSTM-based classifier,the average duration(from the moment starting an attack to the moment the attack being detected)and distance of the proposed scheme are reduced by 24.4%and 19.5%,respectively.展开更多
Visible and infrared(RGB-IR)fusion object detection plays an important role in security,disaster relief,etc.In recent years,deep-learning-based RGB-IR fusion detection methods have been developing rapidly,but still st...Visible and infrared(RGB-IR)fusion object detection plays an important role in security,disaster relief,etc.In recent years,deep-learning-based RGB-IR fusion detection methods have been developing rapidly,but still struggle to deal with the complex and changing scenarios captured by drones,mainly due to two reasons:(A)RGB-IR fusion detectors are susceptible to inferior inputs that degrade performance and stability.(B)RGB-IR fusion detectors are susceptible to redundant features that reduce accuracy and efficiency.In this paper,an innovative RGB-IR fusion detection framework based on global-local feature optimization,named GLFDet,is proposed to improve the detection performance and efficiency of drone-captured objects.The key components of GLFDet include a Global Feature Optimization(GFO)module,a Local Feature Optimization(LFO)module and a Channel Separation Fusion(CSF)module.Specifically,GFO calculates the information content of the input image from the frequency domain and optimizes the features holistically.Then,LFO dynamically selects high-value features and filters out low-value features before fusion,which significantly improves the efficiency of fusion.Finally,CSF fuses the RGB and IR features across the corresponding channels,which avoids the rearrangement of the channel relationships and enhances the model stability.Extensive experimental results show that the proposed method achieves the best performance on three popular RGB-IR datasets Drone Vehicle,VEDAI,and LLVIP.In addition,GLFDet is more lightweight than other comparable models,making it more appealing to edge devices such as drones.The code is available at https://github.com/lao chen330/GLFDet.展开更多
Aiming at the problem that infrared small target detection faces low contrast between the background and the target and insufficient noise suppression ability under the complex cloud background,an infrared small targe...Aiming at the problem that infrared small target detection faces low contrast between the background and the target and insufficient noise suppression ability under the complex cloud background,an infrared small target detection method based on the tensor nuclear norm and direction residual weighting was proposed.Based on converting the infrared image into an infrared patch tensor model,from the perspective of the low-rank nature of the background tensor,and taking advantage of the difference in contrast between the background and the target in different directions,we designed a double-neighborhood local contrast based on direction residual weighting method(DNLCDRW)combined with the partial sum of tensor nuclear norm(PSTNN)to achieve effective background suppression and recovery of infrared small targets.Experiments show that the algorithm is effective in suppressing the background and improving the detection ability of the target.展开更多
The fast increase of online communities has brought about an increase in cyber threats inclusive of cyberbullying, hate speech, misinformation, and online harassment, making content moderation a pressing necessity. Tr...The fast increase of online communities has brought about an increase in cyber threats inclusive of cyberbullying, hate speech, misinformation, and online harassment, making content moderation a pressing necessity. Traditional single-modal AI-based detection systems, which analyze both text, photos, or movies in isolation, have established useless at taking pictures multi-modal threats, in which malicious actors spread dangerous content throughout a couple of formats. To cope with these demanding situations, we advise a multi-modal deep mastering framework that integrates Natural Language Processing (NLP), Convolutional Neural Networks (CNNs), and Long Short-Term Memory (LSTM) networks to become aware of and mitigate online threats effectively. Our proposed model combines BERT for text class, ResNet50 for photograph processing, and a hybrid LSTM-3-d CNN community for video content material analysis. We constructed a large-scale dataset comprising 500,000 textual posts, 200,000 offensive images, and 50,000 annotated motion pictures from more than one platform, which includes Twitter, Reddit, YouTube, and online gaming forums. The system became carefully evaluated using trendy gadget mastering metrics which include accuracy, precision, remember, F1-score, and ROC-AUC curves. Experimental outcomes demonstrate that our multi-modal method extensively outperforms single-modal AI classifiers, achieving an accuracy of 92.3%, precision of 91.2%, do not forget of 90.1%, and an AUC rating of 0.95. The findings validate the necessity of integrating multi-modal AI for actual-time, high-accuracy online chance detection and moderation. Future paintings will have consciousness on improving hostile robustness, enhancing scalability for real-world deployment, and addressing ethical worries associated with AI-driven content moderation.展开更多
With the advancement of deep learning in the automotive domain,more and more researchers are focusing on autonomous driving.Among these tasks,free space detection is particularly crucial.Currently,many model-based app...With the advancement of deep learning in the automotive domain,more and more researchers are focusing on autonomous driving.Among these tasks,free space detection is particularly crucial.Currently,many model-based approaches have achieved autonomous driving on well-structured urban roads,but these efforts primarily focus on urban road environments.In contrast,there are fewer deep learningmethods specifically designed for off-road traversable area detection,and their effectiveness is not yet satisfactory.This is because detecting traversable areas in complex outdoor environments poses significant challenges,and current methods often rely on single-image inputs,which do not align with contemporary multimodal approaches.Therefore,in this study,we propose a CFH-Net model for off-road traversable area detection.This model employs a Transformer architecture to enhance its capability of capturing global information.For multimodal feature extraction and fusion,we integrate the CM-FRM module for feature extraction and introduce the novel FFX module for feature fusion,thereby improving the perception capability of autonomous vehicles on unstructured roads.To address upsampling,we propose a new convolution precorrection method to reduce model parameters and computational complexity while enhancing the model’s ability to capture complex features.Finally,we conducted experiments on the ORFD off-road dataset and achieved outstanding results.展开更多
Infrared small-target detection has important applications in many fields due to its high penetration capability and detection distance.This study introduces a detector called“YOLO-SDLUWD”which is based on the YOLOv...Infrared small-target detection has important applications in many fields due to its high penetration capability and detection distance.This study introduces a detector called“YOLO-SDLUWD”which is based on the YOLOv7 network,for small target detection in complex infrared backgrounds.The“SDLUWD”refers to the combination of the Spatial Depth layer followed Convolutional layer structure(SD-Conv)and a Linear Up-sampling fusion Path Aggregation Feature Pyramid Network(LU-PAFPN)and a training strategy based on the normalized Gaussian Wasserstein Distance loss(WD-loss)function.“YOLO-SDLUWD”aims to reduce detection accuracy when the maximum pooling downsampling layer in the backbone network loses important feature information,support the interaction and fusion of high-dimensional and low-dimensional feature information,and overcome the false alarm predictions induced by noise in small target images.The detector achieved a mAP@0.5 of 90.4%and mAP@0.5:0.95 of 48.5%on IRIS-AG,an increase of 9%-11%over YOLOv7-tiny,outperforming other state-of-the-art target detectors in terms of accuracy and speed.展开更多
基金supported by the Deanship of Graduate Studies and Scientific Research at Jouf University under grant No.(DGSSR-2025-02-01276).
文摘The rapid proliferation of Internet of Things(IoT)devices in critical healthcare infrastructure has introduced significant security and privacy challenges that demand innovative,distributed architectural solutions.This paper proposes FE-ACS(Fog-Edge Adaptive Cybersecurity System),a novel hierarchical security framework that intelligently distributes AI-powered anomaly detection algorithms across edge,fog,and cloud layers to optimize security efficacy,latency,and privacy.Our comprehensive evaluation demonstrates that FE-ACS achieves superior detection performance with an AUC-ROC of 0.985 and an F1-score of 0.923,while maintaining significantly lower end-to-end latency(18.7 ms)compared to cloud-centric(152.3 ms)and fog-only(34.5 ms)architectures.The system exhibits exceptional scalability,supporting up to 38,000 devices with logarithmic performance degradation—a 67×improvement over conventional cloud-based approaches.By incorporating differential privacy mechanisms with balanced privacy-utility tradeoffs(ε=1.0–1.5),FE-ACS maintains 90%–93%detection accuracy while ensuring strong privacy guarantees for sensitive healthcare data.Computational efficiency analysis reveals that our architecture achieves a detection rate of 12,400 events per second with only 12.3 mJ energy consumption per inference.In healthcare risk assessment,FE-ACS demonstrates robust operational viability with low patient safety risk(14.7%)and high system reliability(94.0%).The proposed framework represents a significant advancement in distributed security architectures,offering a scalable,privacy-preserving,and real-time solution for protecting healthcare IoT ecosystems against evolving cyber threats.
文摘Online examinations have become a dominant assessment mode,increasing concerns over academic integrity.To address the critical challenge of detecting cheating behaviours,this study proposes a hybrid deep learning approach that combines visual detection and temporal behaviour classification.The methodology utilises object detection models—You Only Look Once(YOLOv12),Faster Region-based Convolutional Neural Network(RCNN),and Single Shot Detector(SSD)MobileNet—integrated with classification models such as Convolutional Neural Networks(CNN),Bidirectional Gated Recurrent Unit(Bi-GRU),and CNN-LSTM(Long Short-Term Memory).Two distinct datasets were used:the Online Exam Proctoring(EOP)dataset from Michigan State University and the School of Computer Science,Duy Tan Unievrsity(SCS-DTU)dataset collected in a controlled classroom setting.A diverse set of cheating behaviours,including book usage,unauthorised interaction,internet access,and mobile phone use,was categorised.Comprehensive experiments evaluated the models based on accuracy,precision,recall,training time,inference speed,and memory usage.We evaluate nine detector-classifier pairings under a unified budget and score them via a calibrated harmonic mean of detection and classification accuracies,enabling deployment-oriented selection under latency and memory constraints.Macro-Precision/Recall/F1 and Receiver Operating Characteristic-Area Under the Curve(ROC-AUC)are reported for the top configurations,revealing consistent advantages of object-centric pipelines for fine-grained cheating cues.The highest overall score is achieved by YOLOv12+CNN(97.15%accuracy),while SSD-MobileNet+CNN provides the best speed-efficiency trade-off for edge devices.This research provides valuable insights into selecting and deploying appropriate deep learning models for maintaining exam integrity under varying resource constraints.
基金funded by the National Natural Science Foundation of China under Grant 62306128the Leading Innovation Project of Changzhou Science and Technology Bureau underGrant CQ20230072+2 种基金the Basic Science Research Project of Jiangsu Provincial Department of Education under Grant 23KJD520003the Science and Technology Development Plan Project of Jilin Provinceunder Grant 20240101382JCthe National KeyR esearch and Development Program of China under Grant 2023YFF1105102.
文摘In fire rescue scenarios,traditional manual operations are highly dangerous,as dense smoke,low visibility,extreme heat,and toxic gases not only hinder rescue efficiency but also endanger firefighters’safety.Although intelligent rescue robots can enter hazardous environments in place of humans,smoke poses major challenges for human detection algorithms.These challenges include the attenuation of visible and infrared signals,complex thermal fields,and interference frombackground objects,all ofwhichmake it difficult to accurately identify trapped individuals.To address this problem,we propose VIF-YOLO,a visible–infrared fusion model for real-time human detection in dense smoke environments.The framework introduces a lightweight multimodal fusion(LMF)module based on learnable low-rank representation blocks to end-to-end integrate visible and infrared images,preserving fine details while enhancing salient features.In addition,an efficient multiscale attention(EMA)mechanism is incorporated into the YOLOv10n backbone to improve feature representation under low-light conditions.Extensive experiments on our newly constructedmultimodal smoke human detection(MSHD)dataset demonstrate thatVIF-YOLOachievesmAP50 of 99.5%,precision of 99.2%,and recall of 99.3%,outperforming YOLOv10n by a clear margin.Furthermore,when deployed on the NVIDIA Jetson Xavier NX,VIF-YOLO attains 40.6 FPS with an average inference latency of 24.6 ms,validating its real-time capability on edge-computing platforms.These results confirm that VIF-YOLO provides accurate,robust,and fast detection across complex backgrounds and diverse smoke conditions,ensuring reliable and rapid localization of individuals in need of rescue.
基金supported by the National Natural Science Foundation of China(12302056)the Postdoctoral Fellowship Program of China Postdoctoral Science Foundation(GZC20233445)。
文摘Re-entry gliding vehicles exhibit high maneuverability,making trajectory prediction a key factor in the effectiveness of defense systems.To overcome the limited fitting accuracy of existing methods and their poor adaptability to maneuver mode mutations,a trajectory prediction method is proposed that integrates online maneuver mode identification with dynamic modeling.Characteristic parameters are extracted from tracking data for parameterized modeling,enabling real-time identification of maneuver modes.In addition,a maneuver detection mechanism based on higher-order cumulants is introduced to detect lateral maneuver mutations and optimize the use of historical data.Simulation results show that the proposed method achieves accurate trajectory prediction during the glide phase and maintains high accuracy under maneuver mutations,significantly enhancing the prediction performance of both three-dimensional trajectories and ground tracks.
基金supported by the NSFC(Grant Nos.62176273,62271070,62441212)The Open Foundation of State Key Laboratory of Networking and Switching Technology(Beijing University of Posts and Telecommunications)under Grant SKLNST-2024-1-062025Major Project of the Natural Science Foundation of Inner Mongolia(2025ZD008).
文摘The Intrusion Detection System(IDS)is a security mechanism developed to observe network traffic and recognize suspicious or malicious activities.Clustering algorithms are often incorporated into IDS;however,conventional clustering-based methods face notable drawbacks,including poor scalability in handling high-dimensional datasets and a strong dependence of outcomes on initial conditions.To overcome the performance limitations of existing methods,this study proposes a novel quantum-inspired clustering algorithm that relies on a similarity coefficient-based quantum genetic algorithm(SC-QGA)and an improved quantum artificial bee colony algorithm hybrid K-means(IQABC-K).First,the SC-QGA algorithmis constructed based on quantum computing and integrates similarity coefficient theory to strengthen genetic diversity and feature extraction capabilities.For the subsequent clustering phase,the process based on the IQABC-K algorithm is enhanced with the core improvement of adaptive rotation gate and movement exploitation strategies to balance the exploration capabilities of global search and the exploitation capabilities of local search.Simultaneously,the acceleration of convergence toward the global optimum and a reduction in computational complexity are facilitated by means of the global optimum bootstrap strategy and a linear population reduction strategy.Through experimental evaluation with multiple algorithms and diverse performance metrics,the proposed algorithm confirms reliable accuracy on three datasets:KDD CUP99,NSL_KDD,and UNSW_NB15,achieving accuracy of 98.57%,98.81%,and 98.32%,respectively.These results affirm its potential as an effective solution for practical clustering applications.
文摘The problem of fake news detection(FND)is becoming increasingly important in the field of natural language processing(NLP)because of the rapid dissemination of misleading information on the web.Large language models(LLMs)such as GPT-4.Zero excels in natural language understanding tasks but can still struggle to distinguish between fact and fiction,particularly when applied in the wild.However,a key challenge of existing FND methods is that they only consider unimodal data(e.g.,images),while more detailed multimodal data(e.g.,user behaviour,temporal dynamics)is neglected,and the latter is crucial for full-context understanding.To overcome these limitations,we introduce M3-FND(Multimodal Misinformation Mitigation for False News Detection),a novel methodological framework that integrates LLMs with multimodal data sources to perform context-aware veracity assessments.Our method proposes a hybrid system that combines image-text alignment,user credibility profiling,and temporal pattern recognition,which is also strengthened through a natural feedback loop that provides real-time feedback for correcting downstream errors.We use contextual reinforcement learning to schedule prompt updating and update the classifier threshold based on the latest multimodal input,which enables the model to better adapt to changing misinformation attack strategies.M3-FND is tested on three diverse datasets,FakeNewsNet,Twitter15,andWeibo,which contain both text and visual socialmedia content.Experiments showthatM3-FND significantly outperforms conventional and LLMbased baselines in terms of accuracy,F1-score,and AUC on all benchmarks.Our results indicate the importance of employing multimodal cues and adaptive learning for effective and timely detection of fake news.
基金supported by the National Natural Science Foundation of China under Grant 62171233the Natural Science Foundation of China,Jiangsu Province under Grant BK20241891the Jiangsu Province Graduate Research and Practice Innovation Plan under Grants SJCX24_0313 and KYCX24_1169。
文摘A high-sensitivity,low-noise single photon avalanche diode(SPAD)detector was presented based on a 180 nm BCD process.The proposed device utilizes a p-implant layer/high-voltage n-well(HVNW)junction to form a deep avalanche multiplication region for near-infrared(NIR)sensitivity enhancement.By optimizing the device size and electric field of the guard ring,the fill factor(FF)is significantly improved,further increasing photon detection efficiency(PDE).To solve the dark noise caused by the increasing active diameter,a field polysilicon gate structure connected to the p+anode was investigated,effectively suppressing dark count noise by 76.6%.It is experimentally shown that when the active diameter increases from 5 to 10μm,the FF is significantly improved from 20.7%to 39.1%,and thus the peak PDE also rises from 13.3%to 25.8%.At an excess bias voltage of 5 V,a NIR photon detection probability(PDP)of 6.8%at 905 nm,a dark count rate(DCR)of 2.12 cps/μm^(2),an afterpulsing probability(AP)of 1.2%,and a timing jitter of 216 ps are achieved,demonstrating excellent single photon detection performance.
基金supported by National Natural Science Foundation of China(Grant No.:32301161)the Natural Scientific Foundation of Hunan Province,China(Grant No.:2023JJ60052)+3 种基金the Scientific Research Project of Hunan Provincial Health Commission,China(Grant No.:202112062218,20190161)the Scientific Research Project of Hunan Provincial Department of Education,China(Grant No.:22B0455)the Clinical“4310”Project of the University of South China,China(Grant No.:20224310NHYCG02)the Doctoral Scientific Research Foundation of University of South China,China(Grant No.:200XQD042).
文摘Infrared(IR)spectroscopy,a technique within the realm of molecular vibrational spectroscopy,furnishes distinctive chemical signatures pivotal for both structural analysis and compound identification.A notable challenge emerges from the misalignment between the mid-IR light wavelength range and molecular dimensions,culminating in a constrained absorption cross-section and diminished vibrational absorption coefficients(Supplementary data).
文摘Objective:To explore the clinical correlation between the detection of irregular antibodies in red blood cell blood groups and hemolytic disease of the newborn.Methods:This study selected newborns who underwent examinations and were diagnosed with hemolytic disease at our hospital from October 2024 to October 2025 as the research subjects.Based on the severity of their hemolytic disease,the infants were divided into a severe group and a mild group.All the infants underwent detection for irregular antibodies in their red blood cell blood groups.General information,blood types,and irregular antibody test results of the two groups were recorded.Univariate analysis was conducted,and variables with statistical significance from the univariate analysis were included in a multivariate logistic regression analysis to explore the clinical correlation between the detection of irregular antibodies in red blood cell blood groups and hemolytic disease of the newborn.Results:Through univariate analysis,it was found that IgG1 and IgG3 subclass antibodies,as well as ABO blood group incompatibility,were statistically significant(p<0.05).When these factors were included in a multivariate logistic regression analysis,it was discovered that IgG1(OR=2.461,95%CI:1.859-2.709),IgG3(OR=2.509,95%CI:1.918-2.893),and ABO blood group incompatibility(OR=2.998,95%CI:2.149-3.493)all exhibited a positive correlation with hemolytic disease of the newborn.Conclusion:As levels of IgG1,IgG3,and ABO blood group incompatibility increase,the incidence of hemolytic disease of the newborn also rises,warranting clinical attention.
基金supported by grants from the Human Resources Development program (Grant No.20204010600250)the Training Program of CCUS for the Green Growth (Grant No.20214000000500)by the Korea Institute of Energy Technology Evaluation and Planning (KETEP)funded by the Ministry of Trade,Industry,and Energy of the Korean Government (MOTIE).
文摘It is of great importance to obtain precise trace data,as traces are frequently the sole visible and measurable parameter in most outcrops.The manual recognition and detection of traces on high-resolution three-dimensional(3D)models are relatively straightforward but time-consuming.One potential solution to enhance this process is to use machine learning algorithms to detect the 3D traces.In this study,a unique pixel-wise texture mapper algorithm generates a dense point cloud representation of an outcrop with the precise resolution of the original textured 3D model.A virtual digital image rendering was then employed to capture virtual images of selected regions.This technique helps to overcome limitations caused by the surface morphology of the rock mass,such as restricted access,lighting conditions,and shading effects.After AI-powered trace detection on two-dimensional(2D)images,a 3D data structuring technique was applied to the selected trace pixels.In the 3D data structuring,the trace data were structured through 2D thinning,3D reprojection,clustering,segmentation,and segment linking.Finally,the linked segments were exported as 3D polylines,with each polyline in the output corresponding to a trace.The efficacy of the proposed method was assessed using a 3D model of a real-world case study,which was used to compare the results of artificial intelligence(AI)-aided and human intelligence trace detection.Rosette diagrams,which visualize the distribution of trace orientations,confirmed the high similarity between the automatically and manually generated trace maps.In conclusion,the proposed semi-automatic method was easy to use,fast,and accurate in detecting the dominant jointing system of the rock mass.
基金Supported by the Fundamental Research Funds for the Central Universities(2024300443)the Natural Science Foundation of Jiangsu Province(BK20241224).
文摘This paper presents a high-speed and robust dual-band infrared thermal camera based on an ARM CPU.The system consists of a low-resolution long-wavelength infrared detector,a digital temperature and humid⁃ity sensor,and a CMOS sensor.In view of the significant contrast between face and background in thermal infra⁃red images,this paper explores a suitable accuracy-latency tradeoff for thermal face detection and proposes a tiny,lightweight detector named YOLO-Fastest-IR.Four YOLO-Fastest-IR models(IR0 to IR3)with different scales are designed based on YOLO-Fastest.To train and evaluate these lightweight models,a multi-user low-resolution thermal face database(RGBT-MLTF)was collected,and the four networks were trained.Experiments demon⁃strate that the lightweight convolutional neural network performs well in thermal infrared face detection tasks.The proposed algorithm outperforms existing face detection methods in both positioning accuracy and speed,making it more suitable for deployment on mobile platforms or embedded devices.After obtaining the region of interest(ROI)in the infrared(IR)image,the RGB camera is guided by the thermal infrared face detection results to achieve fine positioning of the RGB face.Experimental results show that YOLO-Fastest-IR achieves a frame rate of 92.9 FPS on a Raspberry Pi 4B and successfully detects 97.4%of faces in the RGBT-MLTF test set.Ultimate⁃ly,an infrared temperature measurement system with low cost,strong robustness,and high real-time perfor⁃mance was integrated,achieving a temperature measurement accuracy of 0.3℃.
基金supported by the National Natural Science Foundation of China(No.52472225)the Science and Technology Plan Project of Shenzhen(No.20220808165025003),China。
文摘Visible and near-infrared photodetectors are widely used in intelligent driving,health monitoring,and other fields.However,the application of photodetectors in the near-infrared region is significantly impacted by high dark current,which can greatly reduce their performance and sensitivity,thereby limiting their effectiveness in certain applications.In this work,the introduction of a C60 back interface layer successfully mitigated back interface reactions to decrease the thickness of the Mo(S,Se)_(2)layer,tailoring the back-contact barrier and preventing reverse charge injection,resulting in a kesterite photodetector with an ultralow dark current density of 5.2×10^(-9)mA/cm^(2)and ultra-weak-light detection at levels as low as 25 pW/cm^(2).Besides,under a self-powered operation,it demonstrates outstanding performance,achieving a peak responsivity of 0.68 A/W,a wide response range spanning from 300 to 1600 nm,and an impressive detectivity of 5.27×10^(14)Jones.In addition,it offers exceptionally rapid response times,with rise and decay times of 70 and 650 ns,respectively.This research offers important insights for developing high-performance self-powered near-infrared photodetectors that have high responsivity,rapid response times,and ultralow dark current.
基金supported by Science and Technology Project of State Grid Corporation of China(52094024003D).
文摘As modern power systems grow in complexity,accurate and efficient fault detection has become increasingly important.While many existing reviews focus on a single modality,this paper presents a comprehensive survey from a dual-modality perspective-infrared imaging and voiceprint analysis-two complementary,non-contact techniques that capture different fault characteristics.Infrared imaging excels at detecting thermal anomalies,while voiceprint signals provide insight into mechanical vibrations and internal discharge phenomena.We review both traditional signal processing and deep learning-based approaches for each modality,categorized by key processing stages such as feature extraction and classification.The paper highlights how these modalities address distinct fault types and how they may be fused to improve robustness and accuracy.Representative datasets are summarized,and practical challenges such as noise interference,limited fault samples,and deployment constraints are discussed.By offering a cross-modal,comparative analysis,this work aims to bridge fragmented research and guide future development in intelligent fault detection systems.The review concludes with research trends including multimodal fusion,lightweight models,and self-supervised learning.
文摘In the context of target detection under infrared conditions for drones,the common issues of high missed detection rates,low signal-to-noise ratio,and blurred edge features for small targets are prevalent.To address these challenges,this paper proposes an improved detection algorithm based on YOLOv11n.First,a Dynamic Multi-Scale Feature Fusion and Adaptive Weighting approach is employed to design an Adaptive Focused Diffusion Pyramid Network(AFDPN),which enhances the feature expression and transmission capability of shallow small targets,thereby reducing the loss of detailed information.Then,combined with an Edge Enhancement(EE)module,the model improves the extraction of infrared small target edge features through low-frequency suppression and high-frequency enhancement strategies.Experimental results on the publicly available HIT-UAV dataset show that the improved model achieves a 3.8%increase in average detection accuracy and a 3.0%improvement in recall rate compared to YOLOv11n,with a computational cost of only 9.1 GFLOPS.In comparison experiments,the detection accuracy and model size balance achieved the optimal solution,meeting the lightweight deployment requirements for drone-based systems.This method provides a high-precision,lightweight solution for small target detection in drone-based infrared imagery.
基金supported in part by the National Natural Science Foundation of China(No.62271076)in part by the Fundamental Research Funds for the Central Universities,China(No.2242022k60006).
文摘The Global Positioning System(GPS)plays an indispensable role in the control of Unmanned Aerial Vehicle(UAV).However,the civilian GPS signals,transmitted over the air without any encryption,are vulnerable to spoofing attacks,which further guides the UAV on deviated positions or trajectories.To counter the GPS,,m spoofing on UAV system and to detect the position/trajectory anomaly in real time,a motion state vector based stack long short-term memory trajectory prediction scheme is firstly proposed,leveraging the temporal and spatial features of UAV kinematics.Based on the predicted results,an ensemble voting-based trajectory anomaly detection scheme is proposed to detect the position anomalies in real time with the information of motion state sequences.The proposed prediction-based trajectory anomaly detection scheme outperforms the existing offline detection schemes designed for fixed trajectories.Software In The Loop(SITL)based online prediction and online anomaly detection are demonstrated with random 3D flight trajectories.Results show that the coefficient of determination(R^(2))and Root Mean Square Error(RMSE)of the prediction scheme can reach 0.996 and 3.467,respectively.The accuracy,recall,and F1-score of the proposed anomaly detection scheme can reach 0.984,0.988,and 0.983,respectively,which outperform deep ensemble learning,LSTM-based classifier,machine learning classifier and GA-XGBoost based schemes.Moreover,results show that compared with LSTM-based classifier,the average duration(from the moment starting an attack to the moment the attack being detected)and distance of the proposed scheme are reduced by 24.4%and 19.5%,respectively.
基金supported by the National Natural Science Foundation of China(No.62276204)the Fundamental Research Funds for the Central Universities,China(No.YJSJ24011)+1 种基金the Natural Science Basic Research Program of Shaanxi,China(Nos.2022JM-340 and 2023-JC-QN-0710)the China Postdoctoral Science Foundation(Nos.2020T130494 and 2018M633470)。
文摘Visible and infrared(RGB-IR)fusion object detection plays an important role in security,disaster relief,etc.In recent years,deep-learning-based RGB-IR fusion detection methods have been developing rapidly,but still struggle to deal with the complex and changing scenarios captured by drones,mainly due to two reasons:(A)RGB-IR fusion detectors are susceptible to inferior inputs that degrade performance and stability.(B)RGB-IR fusion detectors are susceptible to redundant features that reduce accuracy and efficiency.In this paper,an innovative RGB-IR fusion detection framework based on global-local feature optimization,named GLFDet,is proposed to improve the detection performance and efficiency of drone-captured objects.The key components of GLFDet include a Global Feature Optimization(GFO)module,a Local Feature Optimization(LFO)module and a Channel Separation Fusion(CSF)module.Specifically,GFO calculates the information content of the input image from the frequency domain and optimizes the features holistically.Then,LFO dynamically selects high-value features and filters out low-value features before fusion,which significantly improves the efficiency of fusion.Finally,CSF fuses the RGB and IR features across the corresponding channels,which avoids the rearrangement of the channel relationships and enhances the model stability.Extensive experimental results show that the proposed method achieves the best performance on three popular RGB-IR datasets Drone Vehicle,VEDAI,and LLVIP.In addition,GLFDet is more lightweight than other comparable models,making it more appealing to edge devices such as drones.The code is available at https://github.com/lao chen330/GLFDet.
基金Supported by the Key Laboratory Fund for Equipment Pre-Research(6142207210202)。
文摘Aiming at the problem that infrared small target detection faces low contrast between the background and the target and insufficient noise suppression ability under the complex cloud background,an infrared small target detection method based on the tensor nuclear norm and direction residual weighting was proposed.Based on converting the infrared image into an infrared patch tensor model,from the perspective of the low-rank nature of the background tensor,and taking advantage of the difference in contrast between the background and the target in different directions,we designed a double-neighborhood local contrast based on direction residual weighting method(DNLCDRW)combined with the partial sum of tensor nuclear norm(PSTNN)to achieve effective background suppression and recovery of infrared small targets.Experiments show that the algorithm is effective in suppressing the background and improving the detection ability of the target.
文摘The fast increase of online communities has brought about an increase in cyber threats inclusive of cyberbullying, hate speech, misinformation, and online harassment, making content moderation a pressing necessity. Traditional single-modal AI-based detection systems, which analyze both text, photos, or movies in isolation, have established useless at taking pictures multi-modal threats, in which malicious actors spread dangerous content throughout a couple of formats. To cope with these demanding situations, we advise a multi-modal deep mastering framework that integrates Natural Language Processing (NLP), Convolutional Neural Networks (CNNs), and Long Short-Term Memory (LSTM) networks to become aware of and mitigate online threats effectively. Our proposed model combines BERT for text class, ResNet50 for photograph processing, and a hybrid LSTM-3-d CNN community for video content material analysis. We constructed a large-scale dataset comprising 500,000 textual posts, 200,000 offensive images, and 50,000 annotated motion pictures from more than one platform, which includes Twitter, Reddit, YouTube, and online gaming forums. The system became carefully evaluated using trendy gadget mastering metrics which include accuracy, precision, remember, F1-score, and ROC-AUC curves. Experimental outcomes demonstrate that our multi-modal method extensively outperforms single-modal AI classifiers, achieving an accuracy of 92.3%, precision of 91.2%, do not forget of 90.1%, and an AUC rating of 0.95. The findings validate the necessity of integrating multi-modal AI for actual-time, high-accuracy online chance detection and moderation. Future paintings will have consciousness on improving hostile robustness, enhancing scalability for real-world deployment, and addressing ethical worries associated with AI-driven content moderation.
文摘With the advancement of deep learning in the automotive domain,more and more researchers are focusing on autonomous driving.Among these tasks,free space detection is particularly crucial.Currently,many model-based approaches have achieved autonomous driving on well-structured urban roads,but these efforts primarily focus on urban road environments.In contrast,there are fewer deep learningmethods specifically designed for off-road traversable area detection,and their effectiveness is not yet satisfactory.This is because detecting traversable areas in complex outdoor environments poses significant challenges,and current methods often rely on single-image inputs,which do not align with contemporary multimodal approaches.Therefore,in this study,we propose a CFH-Net model for off-road traversable area detection.This model employs a Transformer architecture to enhance its capability of capturing global information.For multimodal feature extraction and fusion,we integrate the CM-FRM module for feature extraction and introduce the novel FFX module for feature fusion,thereby improving the perception capability of autonomous vehicles on unstructured roads.To address upsampling,we propose a new convolution precorrection method to reduce model parameters and computational complexity while enhancing the model’s ability to capture complex features.Finally,we conducted experiments on the ORFD off-road dataset and achieved outstanding results.
基金supported by the National Key R&D Program“Development and Application Verification of Underwater Intelligent Defect Detection Robot System for Large Hydropower Station Dams”(Project No.2022YFB4703400)sub-topic 4“Research on Intelligent Identification and Diagnosis of Dam Defects and Fine Inspection Equipment and Technology of Hydropower Stations”(Project No.2022YFB4703404)supported in part by the National Natural Science Foundation of China under Grant 62371181in part by the Changzhou Science and Technology International Cooperation Program under Grant CZ20230029。
文摘Infrared small-target detection has important applications in many fields due to its high penetration capability and detection distance.This study introduces a detector called“YOLO-SDLUWD”which is based on the YOLOv7 network,for small target detection in complex infrared backgrounds.The“SDLUWD”refers to the combination of the Spatial Depth layer followed Convolutional layer structure(SD-Conv)and a Linear Up-sampling fusion Path Aggregation Feature Pyramid Network(LU-PAFPN)and a training strategy based on the normalized Gaussian Wasserstein Distance loss(WD-loss)function.“YOLO-SDLUWD”aims to reduce detection accuracy when the maximum pooling downsampling layer in the backbone network loses important feature information,support the interaction and fusion of high-dimensional and low-dimensional feature information,and overcome the false alarm predictions induced by noise in small target images.The detector achieved a mAP@0.5 of 90.4%and mAP@0.5:0.95 of 48.5%on IRIS-AG,an increase of 9%-11%over YOLOv7-tiny,outperforming other state-of-the-art target detectors in terms of accuracy and speed.