Discriminative region localization and efficient feature encoding are crucial for fine-grained object recognition.However,existing data augmentation methods struggle to accurately locate discriminative regions in comp...Discriminative region localization and efficient feature encoding are crucial for fine-grained object recognition.However,existing data augmentation methods struggle to accurately locate discriminative regions in complex backgrounds,small target objects,and limited training data,leading to poor recognition.Fine-grained images exhibit“small inter-class differences,”and while second-order feature encoding enhances discrimination,it often requires dual Convolutional Neural Networks(CNN),increasing training time and complexity.This study proposes a model integrating discriminative region localization and efficient second-order feature encoding.By ranking feature map channels via a fully connected layer,it selects high-importance channels to generate an enhanced map,accurately locating discriminative regions.Cropping and erasing augmentations further refine recognition.To improve efficiency,a novel second-order feature encoding module generates an attention map from the fourth convolutional group of Residual Network 50 layers(ResNet-50)and multiplies it with features from the fifth group,producing second-order features while reducing dimensionality and training time.Experiments on Caltech-University of California,San Diego Birds-200-2011(CUB-200-2011),Stanford Car,and Fine-Grained Visual Classification of Aircraft(FGVC Aircraft)datasets show state-of-the-art accuracy of 88.9%,94.7%,and 93.3%,respectively.展开更多
Deep learning has made significant progress in the field of oriented object detection for remote sensing images.However,existing methods still face challenges when dealing with difficult tasks such as multi-scale targ...Deep learning has made significant progress in the field of oriented object detection for remote sensing images.However,existing methods still face challenges when dealing with difficult tasks such as multi-scale targets,complex backgrounds,and small objects in remote sensing.Maintaining model lightweight to address resource constraints in remote sensing scenarios while improving task completion for remote sensing tasks remains a research hotspot.Therefore,we propose an enhanced multi-scale feature extraction lightweight network EM-YOLO based on the YOLOv8s architecture,specifically optimized for the characteristics of large target scale variations,diverse orientations,and numerous small objects in remote sensing images.Our innovations lie in two main aspects:First,a dynamic snake convolution(DSC)is introduced into the backbone network to enhance the model’s feature extraction capability for oriented targets.Second,an innovative focusing-diffusion module is designed in the feature fusion neck to effectively integrate multi-scale feature information.Finally,we introduce Layer-Adaptive Sparsity for magnitude-based Pruning(LASP)method to perform lightweight network pruning to better complete tasks in resource-constrained scenarios.Experimental results on the lightweight platform Orin demonstrate that the proposed method significantly outperforms the original YOLOv8s model in oriented remote sensing object detection tasks,and achieves comparable or superior performance to state-of-the-art methods on three authoritative remote sensing datasets(DOTA v1.0,DOTA v1.5,and HRSC2016).展开更多
The rapid development of electricity retail market has prompted an increasing number of electricity consumers to sign green electricity contracts with retail electricity companies,which poses greater challenges for th...The rapid development of electricity retail market has prompted an increasing number of electricity consumers to sign green electricity contracts with retail electricity companies,which poses greater challenges for the market service for green energy consumers.This study proposed a two-stage feature extraction approach for green energy consumers leveraging clustering and termfrequency-inverse document frequency(TF-IDF)algorithms within a knowledge graph framework to provide an information basis that supports the green development of the retail electricity market.First,the multi-source heterogeneous data of green energy consumers under an actual market environment is systematically introduced and the information is categorized into discrete,interval,and relational features.A clustering algorithm was employed to extract features of the trading behavior of green energy consumers in the first stage using the parameter data of green retail electricity contracts.Then,TF-IDF algorithm was applied in the second stage to extract features for green energy consumers in different clusters.Finally,the effectiveness of the proposed approach was validated based on the actual operational data in a southern province of China.It is shown that the most significant discrepancy between the retail trading behaviors of green energy consumers is the power share of green retail packages,whose averaged values are 25.64%,50%,39.66%,and 24.89%in four different clusters,respectively.Additionally,power supply bureaus and electricity retail companies affects the behavior of the green energy consumers most significantly.展开更多
Joint Multimodal Aspect-based Sentiment Analysis(JMASA)is a significant task in the research of multimodal fine-grained sentiment analysis,which combines two subtasks:Multimodal Aspect Term Extraction(MATE)and Multimo...Joint Multimodal Aspect-based Sentiment Analysis(JMASA)is a significant task in the research of multimodal fine-grained sentiment analysis,which combines two subtasks:Multimodal Aspect Term Extraction(MATE)and Multimodal Aspect-oriented Sentiment Classification(MASC).Currently,most existing models for JMASA only perform text and image feature encoding from a basic level,but often neglect the in-depth analysis of unimodal intrinsic features,which may lead to the low accuracy of aspect term extraction and the poor ability of sentiment prediction due to the insufficient learning of intra-modal features.Given this problem,we propose a Text-Image Feature Fine-grained Learning(TIFFL)model for JMASA.First,we construct an enhanced adjacency matrix of word dependencies and adopt graph convolutional network to learn the syntactic structure features for text,which addresses the context interference problem of identifying different aspect terms.Then,the adjective-noun pairs extracted from image are introduced to enable the semantic representation of visual features more intuitive,which addresses the ambiguous semantic extraction problem during image feature learning.Thereby,the model performance of aspect term extraction and sentiment polarity prediction can be further optimized and enhanced.Experiments on two Twitter benchmark datasets demonstrate that TIFFL achieves competitive results for JMASA,MATE and MASC,thus validating the effectiveness of our proposed methods.展开更多
With the rapid expansion of social media,analyzing emotions and their causes in texts has gained significant importance.Emotion-cause pair extraction enables the identification of causal relationships between emotions...With the rapid expansion of social media,analyzing emotions and their causes in texts has gained significant importance.Emotion-cause pair extraction enables the identification of causal relationships between emotions and their triggers within a text,facilitating a deeper understanding of expressed sentiments and their underlying reasons.This comprehension is crucial for making informed strategic decisions in various business and societal contexts.However,recent research approaches employing multi-task learning frameworks for modeling often face challenges such as the inability to simultaneouslymodel extracted features and their interactions,or inconsistencies in label prediction between emotion-cause pair extraction and independent assistant tasks like emotion and cause extraction.To address these issues,this study proposes an emotion-cause pair extraction methodology that incorporates joint feature encoding and task alignment mechanisms.The model consists of two primary components:First,joint feature encoding simultaneously generates features for emotion-cause pairs and clauses,enhancing feature interactions between emotion clauses,cause clauses,and emotion-cause pairs.Second,the task alignment technique is applied to reduce the labeling distance between emotion-cause pair extraction and the two assistant tasks,capturing deep semantic information interactions among tasks.The proposed method is evaluated on a Chinese benchmark corpus using 10-fold cross-validation,assessing key performance metrics such as precision,recall,and F1 score.Experimental results demonstrate that the model achieves an F1 score of 76.05%,surpassing the state-of-the-art by 1.03%.The proposed model exhibits significant improvements in emotion-cause pair extraction(ECPE)and cause extraction(CE)compared to existing methods,validating its effectiveness.This research introduces a novel approach based on joint feature encoding and task alignment mechanisms,contributing to advancements in emotion-cause pair extraction.However,the study’s limitation lies in the data sources,potentially restricting the generalizability of the findings.展开更多
The application of transformer networks and feature fusion models in medical image segmentation has aroused considerable attention within the academic circle.Nevertheless,two main obstacles persist:(1)the restrictions...The application of transformer networks and feature fusion models in medical image segmentation has aroused considerable attention within the academic circle.Nevertheless,two main obstacles persist:(1)the restrictions of the Transformer network in dealing with locally detailed features,and(2)the considerable loss of feature information in current feature fusion modules.To solve these issues,this study initially presents a refined feature extraction approach,employing a double-branch feature extraction network to capture complex multi-scale local and global information from images.Subsequently,we proposed a low-loss feature fusion method-Multi-branch Feature Fusion Enhancement Module(MFFEM),which realizes effective feature fusion with minimal loss.Simultaneously,the cross-layer cross-attention fusion module(CLCA)is adopted to further achieve adequate feature fusion by enhancing the interaction between encoders and decoders of various scales.Finally,the feasibility of our method was verified using the Synapse and ACDC datasets,demonstrating its competitiveness.The average DSC(%)was 83.62 and 91.99 respectively,and the average HD95(mm)was reduced to 19.55 and 1.15 respectively.展开更多
The success of robot-assisted pelvic fracture reduction surgery heavily relies on the accuracy of 3D/3D feature-based registration.This process involves extracting anatomical feature points from pre-operative 3D image...The success of robot-assisted pelvic fracture reduction surgery heavily relies on the accuracy of 3D/3D feature-based registration.This process involves extracting anatomical feature points from pre-operative 3D images which can be challenging because of the complex and variable structure of the pelvis.PointMLP_RegNet,a modified PointMLP,was introduced to address this issue.It retains the feature extraction module of PointMLP but replaces the classification layer with a regression layer to predict the coordinates of feature points instead of conducting regular classification.A flowchart for an automatic feature points extraction method was presented,and a series of experiments was conducted on a clinical pelvic dataset to confirm the accuracy and effectiveness of the method.PointMLP_RegNet extracted feature points more accurately,with 8 out of 10 points showing less than 4 mm errors and the remaining two less than 5 mm.Compared to PointNettt and PointNet,it exhibited higher accuracy,robustness and space efficiency.The proposed method will improve the accuracy of anatomical feature points extraction,enhance intra-operative registration precision and facilitate the widespread clinical application of robot-assisted pelvic fracture reduction.展开更多
Real-time detection of surface defects on cables is crucial for ensuring the safe operation of power systems.However,existing methods struggle with small target sizes,complex backgrounds,low-quality image acquisition,...Real-time detection of surface defects on cables is crucial for ensuring the safe operation of power systems.However,existing methods struggle with small target sizes,complex backgrounds,low-quality image acquisition,and interference from contamination.To address these challenges,this paper proposes the Real-time Cable Defect Detection Network(RC2DNet),which achieves an optimal balance between detection accuracy and computational efficiency.Unlike conventional approaches,RC2DNet introduces a small object feature extraction module that enhances the semantic representation of small targets through feature pyramids,multi-level feature fusion,and an adaptive weighting mechanism.Additionally,a boundary feature enhancement module is designed,incorporating boundary-aware convolution,a novel boundary attention mechanism,and an improved loss function to significantly enhance boundary localization accuracy.Experimental results demonstrate that RC2DNet outperforms state-of-the-art methods in precision,recall,F1-score,mean Intersection over Union(mIoU),and frame rate,enabling real-time and highly accurate cable defect detection in complex backgrounds.展开更多
Relation extraction plays a crucial role in numerous downstream tasks.Dialogue relation extraction focuses on identifying relations between two arguments within a given dialogue.To tackle the problem of low informatio...Relation extraction plays a crucial role in numerous downstream tasks.Dialogue relation extraction focuses on identifying relations between two arguments within a given dialogue.To tackle the problem of low information density in dialogues,methods based on trigger enhancement have been proposed,yielding positive results.However,trigger enhancement faces challenges,which cause suboptimal model performance.First,the proportion of annotated triggers is low in DialogRE.Second,feature representations of triggers and arguments often contain conflicting information.In this paper,we propose a novel Multi-Feature Filtering and Fusion trigger enhancement approach to overcome these limitations.We first obtain representations of arguments,and triggers that contain rich semantic information through attention and gate methods.Then,we design a feature filtering mechanism that eliminates conflicting features in the encoding of trigger prototype representations and their corresponding argument pairs.Additionally,we utilize large language models to create prompts based on Chain-of-Thought and In-context Learning for automated trigger extraction.Experiments show that our model increases the average F1 score by 1.3%in the dialogue relation extraction task.Ablation and case studies confirm the effectiveness of our model.Furthermore,the feature filtering method effectively integrates with other trigger enhancement models,enhancing overall performance and demonstrating its ability to resolve feature conflicts.展开更多
This study proposes a multi-scene smoke detection algorithm based on a multi-feature extraction method to address the problems of varying smoke shapes in different scenes,difficulty in locating and detecting transluce...This study proposes a multi-scene smoke detection algorithm based on a multi-feature extraction method to address the problems of varying smoke shapes in different scenes,difficulty in locating and detecting translucent smoke,and variable smoke scales.First,the convolution module of feature extraction in YOLOv5s backbone network is replaced with asymmetric convolution block re-parameterization convolution to improve the detection of different shapes of smoke.Then,coordinate attention mechanism is introduced in the deeper layer of the backbone network to further improve the localization of translucent smoke.Finally,the detection of smoke at different scales is further improved by using the feature pyramid convolution module instead of the standard convolution module of the feature pyramid in the model.The experimental results demonstrate the feasibility and superiority of the proposed model for multi-scene smoke detection.展开更多
Detecting cyber attacks in networks connected to the Internet of Things(IoT)is of utmost importance because of the growing vulnerabilities in the smart environment.Conventional models,such as Naive Bayes and support v...Detecting cyber attacks in networks connected to the Internet of Things(IoT)is of utmost importance because of the growing vulnerabilities in the smart environment.Conventional models,such as Naive Bayes and support vector machine(SVM),as well as ensemble methods,such as Gradient Boosting and eXtreme gradient boosting(XGBoost),are often plagued by high computational costs,which makes it challenging for them to perform real-time detection.In this regard,we suggested an attack detection approach that integrates Visual Geometry Group 16(VGG16),Artificial Rabbits Optimizer(ARO),and Random Forest Model to increase detection accuracy and operational efficiency in Internet of Things(IoT)networks.In the suggested model,the extraction of features from malware pictures was accomplished with the help of VGG16.The prediction process is carried out by the random forest model using the extracted features from the VGG16.Additionally,ARO is used to improve the hyper-parameters of the random forest model of the random forest.With an accuracy of 96.36%,the suggested model outperforms the standard models in terms of accuracy,F1-score,precision,and recall.The comparative research highlights our strategy’s success,which improves performance while maintaining a lower computational cost.This method is ideal for real-time applications,but it is effective.展开更多
Feature extraction in the optical domain offers a promising low-latency,high-throughput solution.Optical diffraction-based feature extraction operating under a coherent light source can further achieve parallel output...Feature extraction in the optical domain offers a promising low-latency,high-throughput solution.Optical diffraction-based feature extraction operating under a coherent light source can further achieve parallel outputs with low energy consumption.However,it presents significant challenges for maintaining the coherent input,scaling the operation rates beyond 10 GHz,and ensuring the effective extraction of functional configuration simultaneously.We propose an optical feature extraction engine(OFE^(2)),which is composed of a diffraction operator and a data preparation module,powering high-speed feature extraction for both image and temporal series tasks.This OFE^(2)can achieve a core latency of less than 250.5 ps;in addition,it can reach a throughput of 250 GOPS and an efficiency of 2.06 TOPS/W.Supported by the OFE^(2),a novel feature extraction paradigm is emerging,enabling high-speed,low-latency service access for applications in scene recognition,medical assistance,and digital finance.展开更多
Recent advancements in smart-meter technology are transforming traditional power systems into intelligent smart grids.It offers substantial benefits across social,environmental,and economic dimensions.To effectively r...Recent advancements in smart-meter technology are transforming traditional power systems into intelligent smart grids.It offers substantial benefits across social,environmental,and economic dimensions.To effectively realize these advantages,a fine-grained collection and analysis of smart meter data is essential.However,the high dimensionality and volume of such time-series present significant challenges,including increased computational load,data transmission overhead,latency,and complexity in real-time analysis.This study proposes a novel,computationally efficient framework for feature extraction and selection tailored to smart meter time-series data.The approach begins with an extensive offline analysis,where features are derived from multiple domains—time,frequency,and statistical—to capture diverse signal characteristics.Various feature sets are fused and evaluated using robust machine learning classifiers to identify the most informative combinations for automated appliance categorization.The bestperforming fused features set undergoes further refinement using Analysis of Variance(ANOVA)to identify the most discriminative features.The mathematical models,used to compute the selected features,are optimized to extract them with computational efficiency during online processing.Moreover,a notable dimension reduction is secured which facilitates data storage,transmission,and post processing.Onward,a specifically designed LogitBoost(LB)based ensemble of Random Forest base learners is used for an automated classification.The proposed solution demonstrates a high classification accuracy(97.93%)for the case of nine-class problem and dimension reduction(17.33-fold)with minimal front-end computational requirements,making it well-suited for real-world applications in smart grid environments.展开更多
Imaging hyperspectral technology has distinctive advantages of non-destructive and non-contact measurement,and the integration of spectral and spatial data.These characteristics present new methodologies for intellige...Imaging hyperspectral technology has distinctive advantages of non-destructive and non-contact measurement,and the integration of spectral and spatial data.These characteristics present new methodologies for intelligent geological sensing in tunnels and other underground engineering projects.However,the in situ acquisition and rapid classification of hyperspectral images in underground still faces great challenges,including the difficulty in obtaining uniform hyperspectral images and the complexity of deploying sophisticated models on mobile platforms.This study proposes an intelligent lithology identification method based on partition feature extraction of hyperspectral images.Firstly,pixel-level hyperspectral information from representative lithological regions is extracted and fused to obtain rock hyperspectral image partition features.Subsequently,an SG-SNV-PCA-DNN(SSPD)model specifically designed for optimizing rock hyperspectral data,performing spectral dimensionality reduction,and identifying lithology is integrated.In an experimental study involving 3420 hyperspectral images,the SSPD identification model achieved the highest accuracy in the testing set,reaching 98.77%.Moreover,the speed of the SSPD model was found to be 18.5%faster than that of the unprocessed model,with an accuracy improvement of 5.22%.In contrast,the ResNet-101 model,used for point-by-point identification based on non-partitioned features,achieved a maximum accuracy of 97.86%in the testing set.In addition,the partition feature extraction methods significantly reduce computational complexity.An objective evaluation of various models demonstrated that the SSPD model exhibited superior performance,achieving a precision(P)of 99.46%,a recall(R)of 99.44%,and F1 score(F1)of 99.45%.Additionally,a pioneering in situ detection work was carried out in a tunnel using underground hyperspectral imaging technology.展开更多
[Objective]Accurate prediction of tomato growth height is crucial for optimizing production environments in smart farming.However,current prediction methods predominantly rely on empirical,mechanistic,or learning-base...[Objective]Accurate prediction of tomato growth height is crucial for optimizing production environments in smart farming.However,current prediction methods predominantly rely on empirical,mechanistic,or learning-based models that utilize either images data or environmental data.These methods fail to fully leverage multi-modal data to capture the diverse aspects of plant growth comprehensively.[Methods]To address this limitation,a two-stage phenotypic feature extraction(PFE)model based on deep learning algorithm of recurrent neural network(RNN)and long short-term memory(LSTM)was developed.The model integrated environment and plant information to provide a holistic understanding of the growth process,emploied phenotypic and temporal feature extractors to comprehensively capture both types of features,enabled a deeper understanding of the interaction between tomato plants and their environment,ultimately leading to highly accurate predictions of growth height.[Results and Discussions]The experimental results showed the model's ef‐fectiveness:When predicting the next two days based on the past five days,the PFE-based RNN and LSTM models achieved mean absolute percentage error(MAPE)of 0.81%and 0.40%,respectively,which were significantly lower than the 8.00%MAPE of the large language model(LLM)and 6.72%MAPE of the Transformer-based model.In longer-term predictions,the 10-day prediction for 4 days ahead and the 30-day prediction for 12 days ahead,the PFE-RNN model continued to outperform the other two baseline models,with MAPE of 2.66%and 14.05%,respectively.[Conclusions]The proposed method,which leverages phenotypic-temporal collaboration,shows great potential for intelligent,data-driven management of tomato cultivation,making it a promising approach for enhancing the efficiency and precision of smart tomato planting management.展开更多
Aiming at the problem of on-line damage diagnosis in structural health monitoring （SHM）, an algorithm of feature extraction and damage alarming based on auto-regressive moving-average （ARMA） time series analysis i...Aiming at the problem of on-line damage diagnosis in structural health monitoring （SHM）, an algorithm of feature extraction and damage alarming based on auto-regressive moving-average （ARMA） time series analysis is presented. The monitoring data were first modeled as ARMA models, while a principalcomponent matrix derived from the AR coefficients of these models was utilized to establish the Mahalanobisdistance criterion functions. Then, a new damage-sensitive feature index DDSF is proposed. A hypothesis test involving the t-test method is further applied to obtain a decision of damage alarming as the mean value of DDSF had significantly changed after damage. The numerical results of a three-span-girder model shows that the defined index is sensitive to subtle structural damage, and the proposed algorithm can be applied to the on-line damage alarming in SHM.展开更多
[Objective] The aim was to study the feature extraction of stored-grain insects based on ant colony optimization and support vector machine algorithm, and to explore the feasibility of the feature extraction of stored...[Objective] The aim was to study the feature extraction of stored-grain insects based on ant colony optimization and support vector machine algorithm, and to explore the feasibility of the feature extraction of stored-grain insects. [Method] Through the analysis of feature extraction in the image recognition of the stored-grain insects, the recognition accuracy of the cross-validation training model in support vector machine （SVM） algorithm was taken as an important factor of the evaluation principle of feature extraction of stored-grain insects. The ant colony optimization （ACO） algorithm was applied to the automatic feature extraction of stored-grain insects. [Result] The algorithm extracted the optimal feature subspace of seven features from the 17 morphological features, including area and perimeter. The ninety image samples of the stored-grain insects were automatically recognized by the optimized SVM classifier, and the recognition accuracy was over 95%. [Conclusion] The experiment shows that the application of ant colony optimization to the feature extraction of grain insects is practical and feasible.展开更多
Aim To extract harmonic frequencies of helicopter acoustic signal as features for hel icopter identification. Methods Estimation of signal parameters via rotational invariance techniques(ESPRIT) was selected to ext...Aim To extract harmonic frequencies of helicopter acoustic signal as features for hel icopter identification. Methods Estimation of signal parameters via rotational invariance techniques(ESPRIT) was selected to extract harmonic frequencies from really measured helicopter acoustic signal and an algorithm based on the SVD TLS was used. Results ESPRIT correctly extracted harmonic frequencies of helicopter using the data of limited length under the variousflight conditions. Conclusion ESPRIT is an effective method of extracting harmonic frequencies and using harmonic frequencies of helicopter acoustic signal to recognize helicopter is feasible.展开更多
Due to the non-stationary characteristics of vibration signals acquired from rolling element bearing fault, thc time-frequency analysis is often applied to describe the local information of these unstable signals smar...Due to the non-stationary characteristics of vibration signals acquired from rolling element bearing fault, thc time-frequency analysis is often applied to describe the local information of these unstable signals smartly. However, it is difficult to classitythe high dimensional feature matrix directly because of too large dimensions for many classifiers. This paper combines the concepts of time-frequency distribution（TFD） with non-negative matrix factorization（NMF）, and proposes a novel TFD matrix factorization method to enhance representation and identification of bearing fault. Throughout this method, the TFD of a vibration signal is firstly accomplished to describe the localized faults with short-time Fourier transform（STFT）. Then, the supervised NMF mapping is adopted to extract the fault features from TFD. Meanwhile, the fault samples can be clustered and recognized automatically by using the clustering property of NMF. The proposed method takes advantages of the NMF in the parts-based representation and the adaptive clustering. The localized fault features of interest can be extracted as well. To evaluate the performance of the proposed method, the 9 kinds of the bearing fault on a test bench is performed. The proposed method can effectively identify the fault severity and different fault types. Moreover, in comparison with the artificial neural network（ANN）, NMF yields 99.3% mean accuracy which is much superior to ANN. This research presents a simple and practical resolution for the fault diagnosis problem of rolling element bearing in high dimensional feature space.展开更多
Choosing the right characteristic parameter is the key to fault diagnosis in analog circuit. The feature evaluation and extraction methods based on neural network are presented. Parameter evaluation of circuit feature...Choosing the right characteristic parameter is the key to fault diagnosis in analog circuit. The feature evaluation and extraction methods based on neural network are presented. Parameter evaluation of circuit features is realized by training results from neural network; the superior nonlinear mapping capability is competent for extracting fault features which are normalized and compressed subsequently. The complex classification problem on fault pattern recognition in analog circuit is transferred into feature processing stage by feature extraction based on neural network effectively, which improves the diagnosis efficiency. A fault diagnosis illustration validated this method.展开更多
基金supported,in part,by the National Nature Science Foundation of China under Grant 62272236,62376128 and 62306139the Natural Science Foundation of Jiangsu Province under Grant BK20201136,BK20191401.
文摘Discriminative region localization and efficient feature encoding are crucial for fine-grained object recognition.However,existing data augmentation methods struggle to accurately locate discriminative regions in complex backgrounds,small target objects,and limited training data,leading to poor recognition.Fine-grained images exhibit“small inter-class differences,”and while second-order feature encoding enhances discrimination,it often requires dual Convolutional Neural Networks(CNN),increasing training time and complexity.This study proposes a model integrating discriminative region localization and efficient second-order feature encoding.By ranking feature map channels via a fully connected layer,it selects high-importance channels to generate an enhanced map,accurately locating discriminative regions.Cropping and erasing augmentations further refine recognition.To improve efficiency,a novel second-order feature encoding module generates an attention map from the fourth convolutional group of Residual Network 50 layers(ResNet-50)and multiplies it with features from the fifth group,producing second-order features while reducing dimensionality and training time.Experiments on Caltech-University of California,San Diego Birds-200-2011(CUB-200-2011),Stanford Car,and Fine-Grained Visual Classification of Aircraft(FGVC Aircraft)datasets show state-of-the-art accuracy of 88.9%,94.7%,and 93.3%,respectively.
基金funded by the Hainan Province Science and Technology Special Fund under Grant ZDYF2024GXJS292.
文摘Deep learning has made significant progress in the field of oriented object detection for remote sensing images.However,existing methods still face challenges when dealing with difficult tasks such as multi-scale targets,complex backgrounds,and small objects in remote sensing.Maintaining model lightweight to address resource constraints in remote sensing scenarios while improving task completion for remote sensing tasks remains a research hotspot.Therefore,we propose an enhanced multi-scale feature extraction lightweight network EM-YOLO based on the YOLOv8s architecture,specifically optimized for the characteristics of large target scale variations,diverse orientations,and numerous small objects in remote sensing images.Our innovations lie in two main aspects:First,a dynamic snake convolution(DSC)is introduced into the backbone network to enhance the model’s feature extraction capability for oriented targets.Second,an innovative focusing-diffusion module is designed in the feature fusion neck to effectively integrate multi-scale feature information.Finally,we introduce Layer-Adaptive Sparsity for magnitude-based Pruning(LASP)method to perform lightweight network pruning to better complete tasks in resource-constrained scenarios.Experimental results on the lightweight platform Orin demonstrate that the proposed method significantly outperforms the original YOLOv8s model in oriented remote sensing object detection tasks,and achieves comparable or superior performance to state-of-the-art methods on three authoritative remote sensing datasets(DOTA v1.0,DOTA v1.5,and HRSC2016).
基金support by the Science and Technology Project of Guangdong Power Exchange Center Co.,Ltd.(No.GDKJXM20222599)National Natural Science Foundation of China(No.52207104)Natural Science Foundation of Guangdong Province(No.2024A1515010426).
文摘The rapid development of electricity retail market has prompted an increasing number of electricity consumers to sign green electricity contracts with retail electricity companies,which poses greater challenges for the market service for green energy consumers.This study proposed a two-stage feature extraction approach for green energy consumers leveraging clustering and termfrequency-inverse document frequency(TF-IDF)algorithms within a knowledge graph framework to provide an information basis that supports the green development of the retail electricity market.First,the multi-source heterogeneous data of green energy consumers under an actual market environment is systematically introduced and the information is categorized into discrete,interval,and relational features.A clustering algorithm was employed to extract features of the trading behavior of green energy consumers in the first stage using the parameter data of green retail electricity contracts.Then,TF-IDF algorithm was applied in the second stage to extract features for green energy consumers in different clusters.Finally,the effectiveness of the proposed approach was validated based on the actual operational data in a southern province of China.It is shown that the most significant discrepancy between the retail trading behaviors of green energy consumers is the power share of green retail packages,whose averaged values are 25.64%,50%,39.66%,and 24.89%in four different clusters,respectively.Additionally,power supply bureaus and electricity retail companies affects the behavior of the green energy consumers most significantly.
基金supported by the Science and Technology Project of Henan Province(No.222102210081).
文摘Joint Multimodal Aspect-based Sentiment Analysis(JMASA)is a significant task in the research of multimodal fine-grained sentiment analysis,which combines two subtasks:Multimodal Aspect Term Extraction(MATE)and Multimodal Aspect-oriented Sentiment Classification(MASC).Currently,most existing models for JMASA only perform text and image feature encoding from a basic level,but often neglect the in-depth analysis of unimodal intrinsic features,which may lead to the low accuracy of aspect term extraction and the poor ability of sentiment prediction due to the insufficient learning of intra-modal features.Given this problem,we propose a Text-Image Feature Fine-grained Learning(TIFFL)model for JMASA.First,we construct an enhanced adjacency matrix of word dependencies and adopt graph convolutional network to learn the syntactic structure features for text,which addresses the context interference problem of identifying different aspect terms.Then,the adjective-noun pairs extracted from image are introduced to enable the semantic representation of visual features more intuitive,which addresses the ambiguous semantic extraction problem during image feature learning.Thereby,the model performance of aspect term extraction and sentiment polarity prediction can be further optimized and enhanced.Experiments on two Twitter benchmark datasets demonstrate that TIFFL achieves competitive results for JMASA,MATE and MASC,thus validating the effectiveness of our proposed methods.
文摘With the rapid expansion of social media,analyzing emotions and their causes in texts has gained significant importance.Emotion-cause pair extraction enables the identification of causal relationships between emotions and their triggers within a text,facilitating a deeper understanding of expressed sentiments and their underlying reasons.This comprehension is crucial for making informed strategic decisions in various business and societal contexts.However,recent research approaches employing multi-task learning frameworks for modeling often face challenges such as the inability to simultaneouslymodel extracted features and their interactions,or inconsistencies in label prediction between emotion-cause pair extraction and independent assistant tasks like emotion and cause extraction.To address these issues,this study proposes an emotion-cause pair extraction methodology that incorporates joint feature encoding and task alignment mechanisms.The model consists of two primary components:First,joint feature encoding simultaneously generates features for emotion-cause pairs and clauses,enhancing feature interactions between emotion clauses,cause clauses,and emotion-cause pairs.Second,the task alignment technique is applied to reduce the labeling distance between emotion-cause pair extraction and the two assistant tasks,capturing deep semantic information interactions among tasks.The proposed method is evaluated on a Chinese benchmark corpus using 10-fold cross-validation,assessing key performance metrics such as precision,recall,and F1 score.Experimental results demonstrate that the model achieves an F1 score of 76.05%,surpassing the state-of-the-art by 1.03%.The proposed model exhibits significant improvements in emotion-cause pair extraction(ECPE)and cause extraction(CE)compared to existing methods,validating its effectiveness.This research introduces a novel approach based on joint feature encoding and task alignment mechanisms,contributing to advancements in emotion-cause pair extraction.However,the study’s limitation lies in the data sources,potentially restricting the generalizability of the findings.
基金funded by the Henan Science and Technology research project(222103810042)Support by the open project of scientific research platform of grain information processing center of Henan University of Technology(KFJJ-2021-108)+1 种基金Support by the innovative funds plan of Henan University of Technology(2021ZKCJ14)Henan University of Technology Youth Backbone Teacher Program.
文摘The application of transformer networks and feature fusion models in medical image segmentation has aroused considerable attention within the academic circle.Nevertheless,two main obstacles persist:(1)the restrictions of the Transformer network in dealing with locally detailed features,and(2)the considerable loss of feature information in current feature fusion modules.To solve these issues,this study initially presents a refined feature extraction approach,employing a double-branch feature extraction network to capture complex multi-scale local and global information from images.Subsequently,we proposed a low-loss feature fusion method-Multi-branch Feature Fusion Enhancement Module(MFFEM),which realizes effective feature fusion with minimal loss.Simultaneously,the cross-layer cross-attention fusion module(CLCA)is adopted to further achieve adequate feature fusion by enhancing the interaction between encoders and decoders of various scales.Finally,the feasibility of our method was verified using the Synapse and ACDC datasets,demonstrating its competitiveness.The average DSC(%)was 83.62 and 91.99 respectively,and the average HD95(mm)was reduced to 19.55 and 1.15 respectively.
基金supported by the National Key Research and Development Program of China(Grant No.2020YFB1313800)the National Science Foundation of China(Grant No.NSFC62373259)+1 种基金the Natural Science Foundation of Top Talent of SZTU(Grant No.GDRC202303)the Education Promotion Foundation of Guangdong Province(Grant No.2022ZDJS115).
文摘The success of robot-assisted pelvic fracture reduction surgery heavily relies on the accuracy of 3D/3D feature-based registration.This process involves extracting anatomical feature points from pre-operative 3D images which can be challenging because of the complex and variable structure of the pelvis.PointMLP_RegNet,a modified PointMLP,was introduced to address this issue.It retains the feature extraction module of PointMLP but replaces the classification layer with a regression layer to predict the coordinates of feature points instead of conducting regular classification.A flowchart for an automatic feature points extraction method was presented,and a series of experiments was conducted on a clinical pelvic dataset to confirm the accuracy and effectiveness of the method.PointMLP_RegNet extracted feature points more accurately,with 8 out of 10 points showing less than 4 mm errors and the remaining two less than 5 mm.Compared to PointNettt and PointNet,it exhibited higher accuracy,robustness and space efficiency.The proposed method will improve the accuracy of anatomical feature points extraction,enhance intra-operative registration precision and facilitate the widespread clinical application of robot-assisted pelvic fracture reduction.
基金supported by the National Natural Science Foundation of China under Grant 62306128the Basic Science Research Project of Jiangsu Provincial Department of Education under Grant 23KJD520003the Leading Innovation Project of Changzhou Science and Technology Bureau under Grant CQ20230072.
文摘Real-time detection of surface defects on cables is crucial for ensuring the safe operation of power systems.However,existing methods struggle with small target sizes,complex backgrounds,low-quality image acquisition,and interference from contamination.To address these challenges,this paper proposes the Real-time Cable Defect Detection Network(RC2DNet),which achieves an optimal balance between detection accuracy and computational efficiency.Unlike conventional approaches,RC2DNet introduces a small object feature extraction module that enhances the semantic representation of small targets through feature pyramids,multi-level feature fusion,and an adaptive weighting mechanism.Additionally,a boundary feature enhancement module is designed,incorporating boundary-aware convolution,a novel boundary attention mechanism,and an improved loss function to significantly enhance boundary localization accuracy.Experimental results demonstrate that RC2DNet outperforms state-of-the-art methods in precision,recall,F1-score,mean Intersection over Union(mIoU),and frame rate,enabling real-time and highly accurate cable defect detection in complex backgrounds.
基金supported by the National Key Research and Development Program of China(No.2023YFF0905400)the National Natural Science Foundation of China(No.U2341229).
文摘Relation extraction plays a crucial role in numerous downstream tasks.Dialogue relation extraction focuses on identifying relations between two arguments within a given dialogue.To tackle the problem of low information density in dialogues,methods based on trigger enhancement have been proposed,yielding positive results.However,trigger enhancement faces challenges,which cause suboptimal model performance.First,the proportion of annotated triggers is low in DialogRE.Second,feature representations of triggers and arguments often contain conflicting information.In this paper,we propose a novel Multi-Feature Filtering and Fusion trigger enhancement approach to overcome these limitations.We first obtain representations of arguments,and triggers that contain rich semantic information through attention and gate methods.Then,we design a feature filtering mechanism that eliminates conflicting features in the encoding of trigger prototype representations and their corresponding argument pairs.Additionally,we utilize large language models to create prompts based on Chain-of-Thought and In-context Learning for automated trigger extraction.Experiments show that our model increases the average F1 score by 1.3%in the dialogue relation extraction task.Ablation and case studies confirm the effectiveness of our model.Furthermore,the feature filtering method effectively integrates with other trigger enhancement models,enhancing overall performance and demonstrating its ability to resolve feature conflicts.
基金the Natural Science Foundation of Zhejiang Province(Nos.LY20F020015 and LY21F020015)the National Natural Science Foundation of China(Nos.61972121 and 61902099)。
文摘This study proposes a multi-scene smoke detection algorithm based on a multi-feature extraction method to address the problems of varying smoke shapes in different scenes,difficulty in locating and detecting translucent smoke,and variable smoke scales.First,the convolution module of feature extraction in YOLOv5s backbone network is replaced with asymmetric convolution block re-parameterization convolution to improve the detection of different shapes of smoke.Then,coordinate attention mechanism is introduced in the deeper layer of the backbone network to further improve the localization of translucent smoke.Finally,the detection of smoke at different scales is further improved by using the feature pyramid convolution module instead of the standard convolution module of the feature pyramid in the model.The experimental results demonstrate the feasibility and superiority of the proposed model for multi-scene smoke detection.
基金funded by Institutional Fund Projects under grant no.(IFPDP-261-22)。
文摘Detecting cyber attacks in networks connected to the Internet of Things(IoT)is of utmost importance because of the growing vulnerabilities in the smart environment.Conventional models,such as Naive Bayes and support vector machine(SVM),as well as ensemble methods,such as Gradient Boosting and eXtreme gradient boosting(XGBoost),are often plagued by high computational costs,which makes it challenging for them to perform real-time detection.In this regard,we suggested an attack detection approach that integrates Visual Geometry Group 16(VGG16),Artificial Rabbits Optimizer(ARO),and Random Forest Model to increase detection accuracy and operational efficiency in Internet of Things(IoT)networks.In the suggested model,the extraction of features from malware pictures was accomplished with the help of VGG16.The prediction process is carried out by the random forest model using the extracted features from the VGG16.Additionally,ARO is used to improve the hyper-parameters of the random forest model of the random forest.With an accuracy of 96.36%,the suggested model outperforms the standard models in terms of accuracy,F1-score,precision,and recall.The comparative research highlights our strategy’s success,which improves performance while maintaining a lower computational cost.This method is ideal for real-time applications,but it is effective.
基金supported by the National Key Research and Development Program of China(Grant No.2024YFE0203600)the National Natural Science Foundation of China(Grant No.62135009).
文摘Feature extraction in the optical domain offers a promising low-latency,high-throughput solution.Optical diffraction-based feature extraction operating under a coherent light source can further achieve parallel outputs with low energy consumption.However,it presents significant challenges for maintaining the coherent input,scaling the operation rates beyond 10 GHz,and ensuring the effective extraction of functional configuration simultaneously.We propose an optical feature extraction engine(OFE^(2)),which is composed of a diffraction operator and a data preparation module,powering high-speed feature extraction for both image and temporal series tasks.This OFE^(2)can achieve a core latency of less than 250.5 ps;in addition,it can reach a throughput of 250 GOPS and an efficiency of 2.06 TOPS/W.Supported by the OFE^(2),a novel feature extraction paradigm is emerging,enabling high-speed,low-latency service access for applications in scene recognition,medical assistance,and digital finance.
文摘Recent advancements in smart-meter technology are transforming traditional power systems into intelligent smart grids.It offers substantial benefits across social,environmental,and economic dimensions.To effectively realize these advantages,a fine-grained collection and analysis of smart meter data is essential.However,the high dimensionality and volume of such time-series present significant challenges,including increased computational load,data transmission overhead,latency,and complexity in real-time analysis.This study proposes a novel,computationally efficient framework for feature extraction and selection tailored to smart meter time-series data.The approach begins with an extensive offline analysis,where features are derived from multiple domains—time,frequency,and statistical—to capture diverse signal characteristics.Various feature sets are fused and evaluated using robust machine learning classifiers to identify the most informative combinations for automated appliance categorization.The bestperforming fused features set undergoes further refinement using Analysis of Variance(ANOVA)to identify the most discriminative features.The mathematical models,used to compute the selected features,are optimized to extract them with computational efficiency during online processing.Moreover,a notable dimension reduction is secured which facilitates data storage,transmission,and post processing.Onward,a specifically designed LogitBoost(LB)based ensemble of Random Forest base learners is used for an automated classification.The proposed solution demonstrates a high classification accuracy(97.93%)for the case of nine-class problem and dimension reduction(17.33-fold)with minimal front-end computational requirements,making it well-suited for real-world applications in smart grid environments.
基金support from the National Natural Science Foundation of China(Grant Nos.52379103,52279103)the Natural Science Foundation of Shandong Province(Grant No.ZR2023YQ049).
文摘Imaging hyperspectral technology has distinctive advantages of non-destructive and non-contact measurement,and the integration of spectral and spatial data.These characteristics present new methodologies for intelligent geological sensing in tunnels and other underground engineering projects.However,the in situ acquisition and rapid classification of hyperspectral images in underground still faces great challenges,including the difficulty in obtaining uniform hyperspectral images and the complexity of deploying sophisticated models on mobile platforms.This study proposes an intelligent lithology identification method based on partition feature extraction of hyperspectral images.Firstly,pixel-level hyperspectral information from representative lithological regions is extracted and fused to obtain rock hyperspectral image partition features.Subsequently,an SG-SNV-PCA-DNN(SSPD)model specifically designed for optimizing rock hyperspectral data,performing spectral dimensionality reduction,and identifying lithology is integrated.In an experimental study involving 3420 hyperspectral images,the SSPD identification model achieved the highest accuracy in the testing set,reaching 98.77%.Moreover,the speed of the SSPD model was found to be 18.5%faster than that of the unprocessed model,with an accuracy improvement of 5.22%.In contrast,the ResNet-101 model,used for point-by-point identification based on non-partitioned features,achieved a maximum accuracy of 97.86%in the testing set.In addition,the partition feature extraction methods significantly reduce computational complexity.An objective evaluation of various models demonstrated that the SSPD model exhibited superior performance,achieving a precision(P)of 99.46%,a recall(R)of 99.44%,and F1 score(F1)of 99.45%.Additionally,a pioneering in situ detection work was carried out in a tunnel using underground hyperspectral imaging technology.
文摘[Objective]Accurate prediction of tomato growth height is crucial for optimizing production environments in smart farming.However,current prediction methods predominantly rely on empirical,mechanistic,or learning-based models that utilize either images data or environmental data.These methods fail to fully leverage multi-modal data to capture the diverse aspects of plant growth comprehensively.[Methods]To address this limitation,a two-stage phenotypic feature extraction(PFE)model based on deep learning algorithm of recurrent neural network(RNN)and long short-term memory(LSTM)was developed.The model integrated environment and plant information to provide a holistic understanding of the growth process,emploied phenotypic and temporal feature extractors to comprehensively capture both types of features,enabled a deeper understanding of the interaction between tomato plants and their environment,ultimately leading to highly accurate predictions of growth height.[Results and Discussions]The experimental results showed the model's ef‐fectiveness:When predicting the next two days based on the past five days,the PFE-based RNN and LSTM models achieved mean absolute percentage error(MAPE)of 0.81%and 0.40%,respectively,which were significantly lower than the 8.00%MAPE of the large language model(LLM)and 6.72%MAPE of the Transformer-based model.In longer-term predictions,the 10-day prediction for 4 days ahead and the 30-day prediction for 12 days ahead,the PFE-RNN model continued to outperform the other two baseline models,with MAPE of 2.66%and 14.05%,respectively.[Conclusions]The proposed method,which leverages phenotypic-temporal collaboration,shows great potential for intelligent,data-driven management of tomato cultivation,making it a promising approach for enhancing the efficiency and precision of smart tomato planting management.
基金The National High Technology Research and Devel-opment Program of China (863Program) (No2006AA04Z416)the National Natural Science Foundation of China (No50538020)
文摘Aiming at the problem of on-line damage diagnosis in structural health monitoring （SHM）, an algorithm of feature extraction and damage alarming based on auto-regressive moving-average （ARMA） time series analysis is presented. The monitoring data were first modeled as ARMA models, while a principalcomponent matrix derived from the AR coefficients of these models was utilized to establish the Mahalanobisdistance criterion functions. Then, a new damage-sensitive feature index DDSF is proposed. A hypothesis test involving the t-test method is further applied to obtain a decision of damage alarming as the mean value of DDSF had significantly changed after damage. The numerical results of a three-span-girder model shows that the defined index is sensitive to subtle structural damage, and the proposed algorithm can be applied to the on-line damage alarming in SHM.
基金Supported by the National Natural Science Foundation of China(31101085)the Program for Young Core Teachers of Colleges in Henan(2011GGJS-094)the Scientific Research Project for the High Level Talents,North China University of Water Conservancy and Hydroelectric Power~~
文摘[Objective] The aim was to study the feature extraction of stored-grain insects based on ant colony optimization and support vector machine algorithm, and to explore the feasibility of the feature extraction of stored-grain insects. [Method] Through the analysis of feature extraction in the image recognition of the stored-grain insects, the recognition accuracy of the cross-validation training model in support vector machine （SVM） algorithm was taken as an important factor of the evaluation principle of feature extraction of stored-grain insects. The ant colony optimization （ACO） algorithm was applied to the automatic feature extraction of stored-grain insects. [Result] The algorithm extracted the optimal feature subspace of seven features from the 17 morphological features, including area and perimeter. The ninety image samples of the stored-grain insects were automatically recognized by the optimized SVM classifier, and the recognition accuracy was over 95%. [Conclusion] The experiment shows that the application of ant colony optimization to the feature extraction of grain insects is practical and feasible.
文摘Aim To extract harmonic frequencies of helicopter acoustic signal as features for hel icopter identification. Methods Estimation of signal parameters via rotational invariance techniques(ESPRIT) was selected to extract harmonic frequencies from really measured helicopter acoustic signal and an algorithm based on the SVD TLS was used. Results ESPRIT correctly extracted harmonic frequencies of helicopter using the data of limited length under the variousflight conditions. Conclusion ESPRIT is an effective method of extracting harmonic frequencies and using harmonic frequencies of helicopter acoustic signal to recognize helicopter is feasible.
基金Supported by Shaanxi Provincial Overall Innovation Project of Science and Technology,China(Grant No.2013KTCQ01-06)
文摘Due to the non-stationary characteristics of vibration signals acquired from rolling element bearing fault, thc time-frequency analysis is often applied to describe the local information of these unstable signals smartly. However, it is difficult to classitythe high dimensional feature matrix directly because of too large dimensions for many classifiers. This paper combines the concepts of time-frequency distribution（TFD） with non-negative matrix factorization（NMF）, and proposes a novel TFD matrix factorization method to enhance representation and identification of bearing fault. Throughout this method, the TFD of a vibration signal is firstly accomplished to describe the localized faults with short-time Fourier transform（STFT）. Then, the supervised NMF mapping is adopted to extract the fault features from TFD. Meanwhile, the fault samples can be clustered and recognized automatically by using the clustering property of NMF. The proposed method takes advantages of the NMF in the parts-based representation and the adaptive clustering. The localized fault features of interest can be extracted as well. To evaluate the performance of the proposed method, the 9 kinds of the bearing fault on a test bench is performed. The proposed method can effectively identify the fault severity and different fault types. Moreover, in comparison with the artificial neural network（ANN）, NMF yields 99.3% mean accuracy which is much superior to ANN. This research presents a simple and practical resolution for the fault diagnosis problem of rolling element bearing in high dimensional feature space.
基金the National Natural Science Fundation of China (60372001 90407007)the Ph. D. Programs Foundation of Ministry of Education of China (20030614006).
文摘Choosing the right characteristic parameter is the key to fault diagnosis in analog circuit. The feature evaluation and extraction methods based on neural network are presented. Parameter evaluation of circuit features is realized by training results from neural network; the superior nonlinear mapping capability is competent for extracting fault features which are normalized and compressed subsequently. The complex classification problem on fault pattern recognition in analog circuit is transferred into feature processing stage by feature extraction based on neural network effectively, which improves the diagnosis efficiency. A fault diagnosis illustration validated this method.
