The correction of Light Detection and Ranging(LiDAR)intensity data is of great significance for enhancing its application value.However,traditional intensity correction methods based on Terrestrial Laser Scanning(TLS)...The correction of Light Detection and Ranging(LiDAR)intensity data is of great significance for enhancing its application value.However,traditional intensity correction methods based on Terrestrial Laser Scanning(TLS)technology rely on manual site setup to collect intensity training data at different distances and incidence angles,which is noisy and limited in sample quantity,restricting the improvement of model accuracy.To overcome this limitation,this study proposes a fine-grained intensity correction modeling method based on Mobile Laser Scanning(MLS)technology.The method utilizes the continuous scanning characteristics of MLS technology to obtain dense point cloud intensity data at various distances and incidence angles.Then,a fine-grained screening strategy is employed to accurately select distance-intensity and incidence angle-intensity modeling samples.Finally,based on these samples,a high-precision intensity correction model is established through polynomial fitting functions.To verify the effectiveness of the proposed method,comparative experiments were designed,and the MLS modeling method was validated against the traditional TLS modeling method on the same test set.The results show that on Test Set 1,where the distance values vary widely(i.e.,0.1–3 m),the intensity consistency after correction using the MLS modeling method reached 7.692 times the original intensity,while the traditional TLS modeling method only increased to 4.630 times the original intensity.On Test Set 2,where the incidence angle values vary widely(i.e.,0○–80○),the MLS modeling method,although with a relatively smaller advantage,still improved the intensity consistency to 3.937 times the original intensity,slightly better than the TLS modeling method’s 3.413 times.These results demonstrate the significant advantage of the modeling method proposed in this study in enhancing the accuracy of intensity correction models.展开更多
Fine-grained weather forecasting data,i.e.,the grid data with high-resolution,have attracted increasing attention in recent years,especially for some specific applications such as the Winter Olympic Games.Although Eur...Fine-grained weather forecasting data,i.e.,the grid data with high-resolution,have attracted increasing attention in recent years,especially for some specific applications such as the Winter Olympic Games.Although European Centre for Medium-Range Weather Forecasts(ECMWF)provides grid prediction up to 240 hours,the coarse data are unable to meet high requirements of these major events.In this paper,we propose a method,called model residual machine learning(MRML),to generate grid prediction with high-resolution based on high-precision stations forecasting.MRML applies model output machine learning(MOML)for stations forecasting.Subsequently,MRML utilizes these forecasts to improve the quality of the grid data by fitting a machine learning(ML)model to the residuals.We demonstrate that MRML achieves high capability at diverse meteorological elements,specifically,temperature,relative humidity,and wind speed.In addition,MRML could be easily extended to other post-processing methods by invoking different techniques.In our experiments,MRML outperforms the traditional downscaling methods such as piecewise linear interpolation(PLI)on the testing data.展开更多
Accurate fine-grained geospatial scene classification using remote sensing imagery is essential for a wide range of applications.However,existing approaches often rely on manually zooming remote sensing images at diff...Accurate fine-grained geospatial scene classification using remote sensing imagery is essential for a wide range of applications.However,existing approaches often rely on manually zooming remote sensing images at different scales to create typical scene samples.This approach fails to adequately support the fixed-resolution image interpretation requirements in real-world scenarios.To address this limitation,we introduce the million-scale fine-grained geospatial scene classification dataset(MEET),which contains over 1.03 million zoom-free remote sensing scene samples,manually annotated into 80 fine-grained categories.In MEET,each scene sample follows a scene-in-scene layout,where the central scene serves as the reference,and auxiliary scenes provide crucial spatial context for fine-grained classification.Moreover,to tackle the emerging challenge of scene-in-scene classification,we present the context-aware transformer(CAT),a model specifically designed for this task,which adaptively fuses spatial context to accurately classify the scene samples.CAT adaptively fuses spatial context to accurately classify the scene samples by learning attentional features that capture the relationships between the center and auxiliary scenes.Based on MEET,we establish a comprehensive benchmark for fine-grained geospatial scene classification,evaluating CAT against 11 competitive baselines.The results demonstrate that CAT significantly outperforms these baselines,achieving a 1.88%higher balanced accuracy(BA)with the Swin-Large backbone,and a notable 7.87%improvement with the Swin-Huge backbone.Further experiments validate the effectiveness of each module in CAT and show the practical applicability of CAT in the urban functional zone mapping.The source code and dataset will be publicly available at https://jerrywyn.github.io/project/MEET.html.展开更多
Fine-grained sediments are widely distributed and constitute the most abundant component in sedi-mentary systems,thus the research on their genesis and distribution is of great significance.In recent years,fine-graine...Fine-grained sediments are widely distributed and constitute the most abundant component in sedi-mentary systems,thus the research on their genesis and distribution is of great significance.In recent years,fine-grained sediment gravity-flows(FGSGF)have been recognized as an important transportation and depositional mechanism for accumulating thick successions of fine-grained sediments.Through a comprehensive review and synthesis of global research on FGSGF deposition,the characteristics,depositional mechanisms,and distribution patterns of fine-grained sediment gravity-flow deposits(FGSGFD)are discussed,and future research prospects are clarified.In addition to the traditionally recognized low-density turbidity current and muddy debris flow,wave-enhanced gravity flow,low-density muddy hyperpycnal flow,and hypopycnal plumes can all form widely distributed FGSGFD.At the same time,the evolution of FGSGF during transportation can result in transitional and hybrid gravity-flow deposits.The combination of multiple triggering mechanisms promotes the widespread develop-ment of FGSGFD,without temporal and spatial limitations.Different types and concentrations of clay minerals,organic matters,and organo-clay complexes are the keys to controlling the flow transformation of FGSGF from low-concentration turbidity currents to high-concentration muddy debris flows.Further study is needed on the interaction mechanism of FGSGF caused by different initiations,the evolution of FGSGF with the effect of organic-inorganic synergy,and the controlling factors of the distribution pat-terns of FGSGFD.The study of FGSGFD can shed some new light on the formation of widely developed thin-bedded siltstones within shales.At the same time,these insights may broaden the exploration scope of shale oil and gas,which have important geological significances for unconventional shale oil and gas.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
A comparison between deep learning and standalone models in predicting the compaction parameters of soil is presented in this research.One hundred and ninety and fifty-three soil samples were randomly picked up from t...A comparison between deep learning and standalone models in predicting the compaction parameters of soil is presented in this research.One hundred and ninety and fifty-three soil samples were randomly picked up from two hundred and forty-three soil samples to create training and validation datasets,respectively.The performance and accuracy of the models were measured by root mean square error(RMSE),coefficient of determination(R2),Pearson product-moment correlation coefficient(r),mean absolute error(MAE),variance accounted for(VAF),mean absolute percentage error(MAPE),weighted mean absolute percentage error(WMAPE),a20-index,index of scatter(IOS),and index of agreement(IOA).Comparisons between standalone models demonstrate that the model MD 29 in Gaussian process regression(GPR)and model MD 101 in support vector machine(SVM)can achieve over 96%of accuracy in predicting the optimum moisture content(OMC)and maximum dry density(MDD)of soil,and outperformed other standalone models.The comparison between deep learning models shows that the models MD 46 and MD 146 in long short-term memory(LSTM)predict OMC and MDD with higher accuracy than ANN models.However,the LSTM models outperformed the GPR models in predicting the compaction parameters.The sensitivity analysis illustrates that fine content(FC),specific gravity(SG),and liquid limit(LL)highly influence the prediction of compaction parameters.展开更多
Based on recent advancements in shale oil exploration within the Ordos Basin,this study presents a comprehensive investigation of the paleoenvironment,lithofacies assemblages and distribution,depositional mechanisms,a...Based on recent advancements in shale oil exploration within the Ordos Basin,this study presents a comprehensive investigation of the paleoenvironment,lithofacies assemblages and distribution,depositional mechanisms,and reservoir characteristics of shale oil of fine-grained sediment deposition in continental freshwater lacustrine basins,with a focus on the Chang 7_(3) sub-member of Triassic Yanchang Formation.The research integrates a variety of exploration data,including field outcrops,drilling,logging,core samples,geochemical analyses,and flume simulation.The study indicates that:(1)The paleoenvironment of the Chang 7_(3) deposition is characterized by a warm and humid climate,frequent monsoon events,and a large water depth of freshwater lacustrine basin.The paleogeomorphology exhibits an asymmetrical pattern,with steep slopes in the southwest and gentle slopes in the northeast,which can be subdivided into microgeomorphological units,including depressions and ridges in lakebed,as well as ancient channels.(2)The Chang 7_(3) sub-member is characterized by a diverse array of fine-grained sediments,including very fine sandstone,siltstone,mudstone and tuff.These sediments are primarily distributed in thin interbedded and laminated arrangements vertically.The overall grain size of the sandstone predominantly falls below 62.5μm,with individual layer thicknesses of 0.05–0.64 m.The deposits contain intact plant fragments and display various sedimentary structure,such as wavy bedding,inverse-to-normal grading sequence,and climbing ripple bedding,which indicating a depositional origin associated with density flows.(3)Flume simulation experiments have successfully replicated the transport processes and sedimentary characteristics associated with density flows.The initial phase is characterized by a density-velocity differential,resulting in a thicker,coarser sediment layer at the flow front,while the upper layers are thinner and finer in grain size.During the mid-phase,sliding water effects cause the fluid front to rise and facilitate rapid forward transport.This process generates multiple“new fronts”,enabling the long-distance transport of fine-grained sandstones,such as siltstone and argillaceous siltstone,into the center of the lake basin.(4)A sedimentary model primarily controlled by hyperpynal flows was established for the southwestern part of the basin,highlighting that the frequent occurrence of flood events and the steep slope topography in this area are primary controlling factors for the development of hyperpynal flows.(5)Sandstone and mudstone in the Chang 7_(3) sub-member exhibit micro-and nano-scale pore-throat systems,shale oil is present in various lithologies,while the content of movable oil varies considerably,with sandstone exhibiting the highest content of movable oil.(6)The fine-grained sediment complexes formed by multiple episodes of sandstones and mudstones associated with density flow in the Chang 7_(3) formation exhibit characteristics of“overall oil-bearing with differential storage capacity”.The combination of mudstone with low total organic carbon content(TOC)and siltstone is identified as the most favorable exploration target at present.展开更多
In this paper,we propose hierarchical attention dual network(DNet)for fine-grained image classification.The DNet can randomly select pairs of inputs from the dataset and compare the differences between them through hi...In this paper,we propose hierarchical attention dual network(DNet)for fine-grained image classification.The DNet can randomly select pairs of inputs from the dataset and compare the differences between them through hierarchical attention feature learning,which are used simultaneously to remove noise and retain salient features.In the loss function,it considers the losses of difference in paired images according to the intra-variance and inter-variance.In addition,we also collect the disaster scene dataset from remote sensing images and apply the proposed method to disaster scene classification,which contains complex scenes and multiple types of disasters.Compared to other methods,experimental results show that the DNet with hierarchical attention is robust to different datasets and performs better.展开更多
To understand the mechanical and twinning behaviors of a fine-grained extruded Mg-8.0Al-0.1Mn-2.0Ca alloy under high-speed impact,impact tests were carried out using a split Hopkinson pressure bar,and microstructures ...To understand the mechanical and twinning behaviors of a fine-grained extruded Mg-8.0Al-0.1Mn-2.0Ca alloy under high-speed impact,impact tests were carried out using a split Hopkinson pressure bar,and microstructures at strains of 0.05,0.10 and 0.20 were obtained using a series of stop rings manufactured by high-strength steel.The stress response and twinning behavior are closely related to loading direction and applied strain rate.The true stress-true strain curves are s-shaped in extrusion direction(ED)and c-shaped in transverse direction(TD),showing apparent anisotropy,while the yield strength is insensitive to loading direction.Almost identical strain-rate sensitivity is demonstrated by the stress in ED and TD.Interestingly,de-twinning is apparent as the applied strain increases to 0.20,and it is enhanced with increasing the applied strain rate.In contrast,the twin density in ED samples is clearly higher than that in TD samples.By modifying the terms of strain hardening and strain rate hardening in the classical JC model,an optimized model is built,which can accurately predict the stress response behavior of the studied alloy under high-speed impact along ED and TD.The correlation coefficient(R)and average absolute relative error(AARE)are 98.63%and 0.0199 for ED,and 96.88%and 0.0202 for TD,respectively.展开更多
Soil responds to cavity expansion is inherently rate-dependent,especially in the case of fine-grained soils.To better understand such rate effects,self-boring pressuremeter tests were conducted on Kunming peaty soil w...Soil responds to cavity expansion is inherently rate-dependent,especially in the case of fine-grained soils.To better understand such rate effects,self-boring pressuremeter tests were conducted on Kunming peaty soil within a strain rate range of 0.1%/min to 5.0%/min.The results showed a clear dependence of cavity pressure and excess pore pressure(EPP)on strain ratesdboth increased with higher rates for a given radial displacement.In light of the experimental results,three cases of cylindrical cavity expansion were investigated using the finite element method and analytical method,partially drained expansion in Modified Cam-Clay(MCC)soil,and undrained and partially drained expansion in elastoviscoplastic(EVP)soil.The EVP behavior was and modeled using the MCC model and the overstress viscoplastic theory.The results indicated that over the strain rate range of 0.0001%/min and 50%/min,the rate response of cavity pressure for the case of partially drained expansion in MCC soil(permeability coefficient ranging from 5×10^(-6) m/s to 2.5×10^(-11) m/s)is not obvious,while the EPP response during undrained expansion in EVP soil shows rate-independent.Only the partially drained solution for cavity expansion in EVP soil captured the rate-sensitive responses of both cavity pressure and EPP,confirmed by the pressuremeter tests on the Kunming peaty soil,Saint-Herblain clay,and Burswood clay.This suggests that the rate effect results from a combination of drainage-related and time-dependent soil behavior.Parametric studies further demonstrated that both viscous behavior and the overconsolidation ratio significantly influence cylindrical cavity expansion response,and the drainage conditions during expansion can be assessed using a nondimensional velocity.展开更多
Climate change is a global phenomenon that has profound impacts on ecological dynamics and biodiversity,shaping the interactions between species and their environment.To gain a deeper understanding of the mechanisms d...Climate change is a global phenomenon that has profound impacts on ecological dynamics and biodiversity,shaping the interactions between species and their environment.To gain a deeper understanding of the mechanisms driving climate change,phenological monitoring is essential.Traditional methods of defining phenological phases often rely on fixed thresholds.However,with the development of technology,deep learning-based classification models are now able to more accurately delineate phenological phases from images,enabling phenological monitoring.Despite the significant advancements these models have made in phenological monitoring,they still face challenges in fully capturing the complexity of biotic-environmental interactions,which can limit the fine-grained accuracy of phenological phase identification.To address this,we propose a novel deep learning model,RESformer,designed to monitor tree phenology at a fine-grained level using PhenoCam images.RESformer features a lightweight structure,making it suitable for deployment in resource-constrained environments.It incorporates a dual-branch routing mechanism that considers both global and local information,thereby improving the accuracy of phenological monitoring.To validate the effectiveness of RESformer,we conducted a case study involving 82,118 images taken over two years from four different locations in Wisconsin,focusing on the phenology of Acer.The images were classified into seven distinct phenological stages,with RESformer achieving an overall monitoring accuracy of 96.02%.Furthermore,we compared RESformer with a phenological monitoring approach based on the Green Chromatic Coordinate(GCC)index and ten popular classification models.The results showed that RESformer excelled in fine-grained monitoring,effectively capturing and identifying changes in phenological stages.This finding not only provides strong support for monitoring the phenology of Acer species but also offers valuable insights for understanding ecological trends and developing more effective ecosystem conservation and management strategies.展开更多
The spray-deposition was used to produce billets of Mg-4Al-1.5Zn-3Ca-1Nd(A alloy)and Mg-13Al-3Zn-3Ca-1Nd(B alloy),and evolution of deformation substructure and Mg_(x)Zn_(y)Ca_(z)metastable phase in fine-grained(3μm)M...The spray-deposition was used to produce billets of Mg-4Al-1.5Zn-3Ca-1Nd(A alloy)and Mg-13Al-3Zn-3Ca-1Nd(B alloy),and evolution of deformation substructure and Mg_(x)Zn_(y)Ca_(z)metastable phase in fine-grained(3μm)Mg alloys was investigated by scanning electron microscopy(SEM),transmission electron microscopy(TEM),X-ray diffraction(XRD),and electron backscattered diffraction(EBSD).It was found that different dislocation configurations were formed in A and B alloys.Redundant free dislocations(RFDs)and dislocation tangles were the ways to form deformation substructure in A alloy,no RFDs except dislocation tangles were found in B alloy.The interaction between nano-scale second phase particles(nano-scale C15 andβ-Mg_(17)(Al,Zn)_(12)phase)and different dislocation configurations had a significant effect on the deformation substructures formation.The mass transfer of Mg_(x)Zn_(y)Ca_(z)metastable phases and the stacking order of stacking faults were conducive to the Mg-Nd-Zn typed long period stacking ordered(LPSO)phases formation.Nano-scale C15 phases,Mg-Nd-Zn typed LPSO phases,c/a ratio,β-Mg_(17)(Al,Zn)_(12)phases were the key factors influencing the formation of textures.Different textures and grain boundary features(GB features)had a significant effect on k-value.The non-basal textures were the main factor affecting k-value in A alloy,while the high-angle grain boundary(HAGB)was the main factor affecting k-value in B alloy.展开更多
Bird monitoring and protection are essential for maintaining biodiversity,and fine-grained bird classification has become a key focus in this field.Audio-visual modalities provide critical cues for this task,but robus...Bird monitoring and protection are essential for maintaining biodiversity,and fine-grained bird classification has become a key focus in this field.Audio-visual modalities provide critical cues for this task,but robust feature extraction and efficient fusion remain major challenges.We introduce a multi-stage fine-grained audiovisual fusion network(MSFG-AVFNet) for fine-grained bird species classification,which addresses these challenges through two key components:(1) the audiovisual feature extraction module,which adopts a multi-stage finetuning strategy to provide high-quality unimodal features,laying a solid foundation for modality fusion;(2) the audiovisual feature fusion module,which combines a max pooling aggregation strategy with a novel audiovisual loss function to achieve effective and robust feature fusion.Experiments were conducted on the self-built AVB81and the publicly available SSW60 datasets,which contain data from 81 and 60 bird species,respectively.Comprehensive experiments demonstrate that our approach achieves notable performance gains,outperforming existing state-of-the-art methods.These results highlight its effectiveness in leveraging audiovisual modalities for fine-grained bird classification and its potential to support ecological monitoring and biodiversity research.展开更多
Fine-grained aircraft target detection in remote sensing holds significant research valueand practical applications,particularly in military defense and precision strikes.Given the complex-ity of remote sensing images...Fine-grained aircraft target detection in remote sensing holds significant research valueand practical applications,particularly in military defense and precision strikes.Given the complex-ity of remote sensing images,where targets are often small and similar within categories,detectingthese fine-grained targets is challenging.To address this,we constructed a fine-grained dataset ofremotely sensed airplanes;for the problems of remote sensing fine-grained targets with obvious head-to-tail distributions and large variations in target sizes,we proposed the DWDet fine-grained tar-get detection and recognition algorithm.First,for the problem of unbalanced category distribution,we adopt an adaptive sampling strategy.In addition,we construct a deformable convolutional blockand improve the decoupling head structure to improve the detection effect of the model ondeformed targets.Then,we design a localization loss function,which is used to improve the model’slocalization ability for targets of different scales.The experimental results show that our algorithmimproves the overall accuracy of the model by 4.1%compared to the baseline model,and improvesthe detection accuracy of small targets by 12.2%.The ablation and comparison experiments alsoprove the effectiveness of our algorithm.展开更多
Fine-grained Image Recognition(FGIR)task is dedicated to distinguishing similar sub-categories that belong to the same super-category,such as bird species and car types.In order to highlight visual differences,existin...Fine-grained Image Recognition(FGIR)task is dedicated to distinguishing similar sub-categories that belong to the same super-category,such as bird species and car types.In order to highlight visual differences,existing FGIR works often follow two steps:discriminative sub-region localization and local feature representation.However,these works pay less attention on global context information.They neglect a fact that the subtle visual difference in challenging scenarios can be highlighted through exploiting the spatial relationship among different subregions from a global view point.Therefore,in this paper,we consider both global and local information for FGIR,and propose a collaborative teacher-student strategy to reinforce and unity the two types of information.Our framework is implemented mainly by convolutional neural network,referred to Teacher-Student Based Attention Convolutional Neural Network(T-S-ACNN).For fine-grained local information,we choose the classic Multi-Attention Network(MA-Net)as our baseline,and propose a type of boundary constraint to further reduce background noises in the local attention maps.In this way,the discriminative sub-regions tend to appear in the area occupied by fine-grained objects,leading to more accurate sub-region localization.For fine-grained global information,we design a graph convolution based Global Attention Network(GA-Net),which can combine extracted local attention maps from MA-Net with non-local techniques to explore spatial relationship among subregions.At last,we develop a collaborative teacher-student strategy to adaptively determine the attended roles and optimization modes,so as to enhance the cooperative reinforcement of MA-Net and GA-Net.Extensive experiments on CUB-200-2011,Stanford Cars and FGVC Aircraft datasets illustrate the promising performance of our framework.展开更多
Coralline soils,specialized materials found extensively in the South China Sea,are playing an increasingly vital role in engineering projects.However,like most terrigenous soils,fine-grained coral soil is prone to shr...Coralline soils,specialized materials found extensively in the South China Sea,are playing an increasingly vital role in engineering projects.However,like most terrigenous soils,fine-grained coral soil is prone to shrinkage and cracking,which can significantly affect its engineering properties and ultimately jeopardize engineering safety.This paper presents a desiccation cracking test of fine-grained coral soil,with a particular focus on the thickness effect.The study involved measuring the water content and recording the evolution of desiccation cracking.Advanced image processing technology is employed to analyze the variations in crack parameters,clod parameters,fractal dimensions,frequency distributions,and desiccation cracking propagation velocities of fine-grained coral soil.Furthermore,the dynamic evolution of desiccation cracking under the influence of layer thickness is analyzed.A comprehensive crack evolution model is proposed,encompassing both top-down and bottom-up crack propagation,as well as internal tensile cracking.This work introduces novel metrics for the propagation velocity of the total crack area,the characteristic propagation velocities of desiccation cracks,and the acceleration of crack propagation.Through data fitting,theoretical formulas for soil water evaporation,propagation velocities of desiccation cracks,and crack propagation acceleration are derived,laying a foundation for future soil cracking theories.展开更多
Although sentiment analysis is pivotal to understanding user preferences,existing models face significant challenges in handling context-dependent sentiments,sarcasm,and nuanced emotions.This study addresses these cha...Although sentiment analysis is pivotal to understanding user preferences,existing models face significant challenges in handling context-dependent sentiments,sarcasm,and nuanced emotions.This study addresses these challenges by integrating ontology-based methods with deep learning models,thereby enhancing sentiment analysis accuracy in complex domains such as film reviews and restaurant feedback.The framework comprises explicit topic recognition,followed by implicit topic identification to mitigate topic interference in subsequent sentiment analysis.In the context of sentiment analysis,we develop an expanded sentiment lexicon based on domainspecific corpora by leveraging techniques such as word-frequency analysis and word embedding.Furthermore,we introduce a sentiment recognition method based on both ontology-derived sentiment features and sentiment lexicons.We evaluate the performance of our system using a dataset of 10,500 restaurant reviews,focusing on sentiment classification accuracy.The incorporation of specialized lexicons and ontology structures enables the framework to discern subtle sentiment variations and context-specific expressions,thereby improving the overall sentiment-analysis performance.Experimental results demonstrate that the integration of ontology-based methods and deep learning models significantly improves sentiment analysis accuracy.展开更多
The fingerprinting-based approach using the wireless local area network(WLAN)is widely used for indoor localization.However,the construction of the fingerprint database is quite time-consuming.Especially when the posi...The fingerprinting-based approach using the wireless local area network(WLAN)is widely used for indoor localization.However,the construction of the fingerprint database is quite time-consuming.Especially when the position of the access point(AP)or wall changes,updating the fingerprint database in real-time is difficult.An appropriate indoor localization approach,which has a low implementation cost,excellent real-time performance,and high localization accuracy and fully considers complex indoor environment factors,is preferred in location-based services(LBSs)applications.In this paper,we proposed a fine-grained grid computing(FGGC)model to achieve decimeter-level localization accuracy.Reference points(RPs)are generated in the grid by the FGGC model.Then,the received signal strength(RSS)values at each RP are calculated with the attenuation factors,such as the frequency band,three-dimensional propagation distance,and walls in complex environments.As a result,the fingerprint database can be established automatically without manual measurement,and the efficiency and cost that the FGGC model takes for the fingerprint database are superior to previous methods.The proposed indoor localization approach,which estimates the position step by step from the approximate grid location to the fine-grained location,can achieve higher real-time performance and localization accuracy simultaneously.The mean error of the proposed model is 0.36 m,far lower than that of previous approaches.Thus,the proposed model is feasible to improve the efficiency and accuracy of Wi-Fi indoor localization.It also shows high-accuracy performance with a fast running speed even under a large-size grid.The results indicate that the proposed method can also be suitable for precise marketing,indoor navigation,and emergency rescue.展开更多
Based on sedimentary characteristics of the fine-grained rocks of the lower submember of second member of the Lower Cretaceous Shahezi Formation(K_(1)sh_(2)^(L))in the Lishu rift depression,combined with methods of or...Based on sedimentary characteristics of the fine-grained rocks of the lower submember of second member of the Lower Cretaceous Shahezi Formation(K_(1)sh_(2)^(L))in the Lishu rift depression,combined with methods of organic petrology,analysis of major and trace elements as well as biological marker compound,the enrichment conditions and enrichment model of organic matter in the fine-grained sedimentary rocks in volcanic rift lacustrine basin are investigated.The change of sedimentary paleoenvironment controls the vertical distribution of different lithofacies types in the K_(1)sh_(2)^(L)and divides it into the upper and lower parts.The lower part contains massive siliceous mudstone with bioclast-bearing siliceous mudstone,whereas the upper part is mostly composed of laminated siliceous shale and laminated fine-grained mixed shale.The kerogen types of organic matter in the lower and upper parts are typesⅡ_(2)–Ⅲand typesⅠ–Ⅱ_(1),respectively.The organic carbon content in the upper part is higher than that in the lower part generally.The enrichment of organic matter in volcanic rift lacustrine basin is subjected to three favorable conditions.First,continuous enhancement of rifting is the direct factor increasing the paleo-water depth,and the rise of base level leads to the expansion of deep-water mudstone/shale deposition range.Second,relatively strong underwater volcanic eruption and rifting are simultaneous,and such event can provide a lot of nutrients for the lake basin,which is conducive to the bloom of algae,resulting in higher productivity of typesⅠ–Ⅱ_(1)kerogen.Third,the relatively dry paleoclimate leads to a decrease in input of fresh water and terrestrial materials,including TypeⅢkerogen from terrestrial higher plants,resulting in a water body with higher salinity and anoxic stratification,which is more favorable for preservation of organic matter.The organic matter enrichment model of fine-grained sedimentary rocks of volcanic rift lacustrine basin is established,which is of reference significance to the understanding of the organic matter enrichment mechanism of fine-grained sedimentary rocks of Shahezi Formation in Songliao Basin and even in the northeast China.展开更多
With the widespread use of SMS(Short Message Service),the proliferation of malicious SMS has emerged as a pressing societal issue.While deep learning-based text classifiers offer promise,they often exhibit suboptimal ...With the widespread use of SMS(Short Message Service),the proliferation of malicious SMS has emerged as a pressing societal issue.While deep learning-based text classifiers offer promise,they often exhibit suboptimal performance in fine-grained detection tasks,primarily due to imbalanced datasets and insufficient model representation capabilities.To address this challenge,this paper proposes an LLMs-enhanced graph fusion dual-stream Transformer model for fine-grained Chinese malicious SMS detection.During the data processing stage,Large Language Models(LLMs)are employed for data augmentation,mitigating dataset imbalance.In the data input stage,both word-level and character-level features are utilized as model inputs,enhancing the richness of features and preventing information loss.A dual-stream Transformer serves as the backbone network in the learning representation stage,complemented by a graph-based feature fusion mechanism.At the output stage,both supervised classification cross-entropy loss and supervised contrastive learning loss are used as multi-task optimization objectives,further enhancing the model’s feature representation.Experimental results demonstrate that the proposed method significantly outperforms baselines on a publicly available Chinese malicious SMS dataset.展开更多
基金supported in part by the National Natural Science Foundation of China under grant number 31901239funded by Researchers Supporting Project Number(RSPD2025R947),King Saud University,Riyadh,Saudi Arabia.
文摘The correction of Light Detection and Ranging(LiDAR)intensity data is of great significance for enhancing its application value.However,traditional intensity correction methods based on Terrestrial Laser Scanning(TLS)technology rely on manual site setup to collect intensity training data at different distances and incidence angles,which is noisy and limited in sample quantity,restricting the improvement of model accuracy.To overcome this limitation,this study proposes a fine-grained intensity correction modeling method based on Mobile Laser Scanning(MLS)technology.The method utilizes the continuous scanning characteristics of MLS technology to obtain dense point cloud intensity data at various distances and incidence angles.Then,a fine-grained screening strategy is employed to accurately select distance-intensity and incidence angle-intensity modeling samples.Finally,based on these samples,a high-precision intensity correction model is established through polynomial fitting functions.To verify the effectiveness of the proposed method,comparative experiments were designed,and the MLS modeling method was validated against the traditional TLS modeling method on the same test set.The results show that on Test Set 1,where the distance values vary widely(i.e.,0.1–3 m),the intensity consistency after correction using the MLS modeling method reached 7.692 times the original intensity,while the traditional TLS modeling method only increased to 4.630 times the original intensity.On Test Set 2,where the incidence angle values vary widely(i.e.,0○–80○),the MLS modeling method,although with a relatively smaller advantage,still improved the intensity consistency to 3.937 times the original intensity,slightly better than the TLS modeling method’s 3.413 times.These results demonstrate the significant advantage of the modeling method proposed in this study in enhancing the accuracy of intensity correction models.
基金Project supported by the National Natural Science Foundation of China(Nos.12101072 and 11421101)the National Key Research and Development Program of China(No.2018YFF0300104)+1 种基金the Beijing Municipal Science and Technology Project(No.Z201100005820002)the Open Research Fund of Shenzhen Research Institute of Big Data(No.2019ORF01001)。
文摘Fine-grained weather forecasting data,i.e.,the grid data with high-resolution,have attracted increasing attention in recent years,especially for some specific applications such as the Winter Olympic Games.Although European Centre for Medium-Range Weather Forecasts(ECMWF)provides grid prediction up to 240 hours,the coarse data are unable to meet high requirements of these major events.In this paper,we propose a method,called model residual machine learning(MRML),to generate grid prediction with high-resolution based on high-precision stations forecasting.MRML applies model output machine learning(MOML)for stations forecasting.Subsequently,MRML utilizes these forecasts to improve the quality of the grid data by fitting a machine learning(ML)model to the residuals.We demonstrate that MRML achieves high capability at diverse meteorological elements,specifically,temperature,relative humidity,and wind speed.In addition,MRML could be easily extended to other post-processing methods by invoking different techniques.In our experiments,MRML outperforms the traditional downscaling methods such as piecewise linear interpolation(PLI)on the testing data.
基金supported by the National Natural Science Foundation of China(42030102,42371321).
文摘Accurate fine-grained geospatial scene classification using remote sensing imagery is essential for a wide range of applications.However,existing approaches often rely on manually zooming remote sensing images at different scales to create typical scene samples.This approach fails to adequately support the fixed-resolution image interpretation requirements in real-world scenarios.To address this limitation,we introduce the million-scale fine-grained geospatial scene classification dataset(MEET),which contains over 1.03 million zoom-free remote sensing scene samples,manually annotated into 80 fine-grained categories.In MEET,each scene sample follows a scene-in-scene layout,where the central scene serves as the reference,and auxiliary scenes provide crucial spatial context for fine-grained classification.Moreover,to tackle the emerging challenge of scene-in-scene classification,we present the context-aware transformer(CAT),a model specifically designed for this task,which adaptively fuses spatial context to accurately classify the scene samples.CAT adaptively fuses spatial context to accurately classify the scene samples by learning attentional features that capture the relationships between the center and auxiliary scenes.Based on MEET,we establish a comprehensive benchmark for fine-grained geospatial scene classification,evaluating CAT against 11 competitive baselines.The results demonstrate that CAT significantly outperforms these baselines,achieving a 1.88%higher balanced accuracy(BA)with the Swin-Large backbone,and a notable 7.87%improvement with the Swin-Huge backbone.Further experiments validate the effectiveness of each module in CAT and show the practical applicability of CAT in the urban functional zone mapping.The source code and dataset will be publicly available at https://jerrywyn.github.io/project/MEET.html.
基金supported by National Natural Science Foundation of China(Grant Nos.42072126,42372139)the Natural Science Foundation of Sichuan Province(Grant Nos.2022NSFSC0990).
文摘Fine-grained sediments are widely distributed and constitute the most abundant component in sedi-mentary systems,thus the research on their genesis and distribution is of great significance.In recent years,fine-grained sediment gravity-flows(FGSGF)have been recognized as an important transportation and depositional mechanism for accumulating thick successions of fine-grained sediments.Through a comprehensive review and synthesis of global research on FGSGF deposition,the characteristics,depositional mechanisms,and distribution patterns of fine-grained sediment gravity-flow deposits(FGSGFD)are discussed,and future research prospects are clarified.In addition to the traditionally recognized low-density turbidity current and muddy debris flow,wave-enhanced gravity flow,low-density muddy hyperpycnal flow,and hypopycnal plumes can all form widely distributed FGSGFD.At the same time,the evolution of FGSGF during transportation can result in transitional and hybrid gravity-flow deposits.The combination of multiple triggering mechanisms promotes the widespread develop-ment of FGSGFD,without temporal and spatial limitations.Different types and concentrations of clay minerals,organic matters,and organo-clay complexes are the keys to controlling the flow transformation of FGSGF from low-concentration turbidity currents to high-concentration muddy debris flows.Further study is needed on the interaction mechanism of FGSGF caused by different initiations,the evolution of FGSGF with the effect of organic-inorganic synergy,and the controlling factors of the distribution pat-terns of FGSGFD.The study of FGSGFD can shed some new light on the formation of widely developed thin-bedded siltstones within shales.At the same time,these insights may broaden the exploration scope of shale oil and gas,which have important geological significances for unconventional shale oil and gas.
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A comparison between deep learning and standalone models in predicting the compaction parameters of soil is presented in this research.One hundred and ninety and fifty-three soil samples were randomly picked up from two hundred and forty-three soil samples to create training and validation datasets,respectively.The performance and accuracy of the models were measured by root mean square error(RMSE),coefficient of determination(R2),Pearson product-moment correlation coefficient(r),mean absolute error(MAE),variance accounted for(VAF),mean absolute percentage error(MAPE),weighted mean absolute percentage error(WMAPE),a20-index,index of scatter(IOS),and index of agreement(IOA).Comparisons between standalone models demonstrate that the model MD 29 in Gaussian process regression(GPR)and model MD 101 in support vector machine(SVM)can achieve over 96%of accuracy in predicting the optimum moisture content(OMC)and maximum dry density(MDD)of soil,and outperformed other standalone models.The comparison between deep learning models shows that the models MD 46 and MD 146 in long short-term memory(LSTM)predict OMC and MDD with higher accuracy than ANN models.However,the LSTM models outperformed the GPR models in predicting the compaction parameters.The sensitivity analysis illustrates that fine content(FC),specific gravity(SG),and liquid limit(LL)highly influence the prediction of compaction parameters.
基金Supported by the CNPC Major Science and Technology Project(2021DJ1806).
文摘Based on recent advancements in shale oil exploration within the Ordos Basin,this study presents a comprehensive investigation of the paleoenvironment,lithofacies assemblages and distribution,depositional mechanisms,and reservoir characteristics of shale oil of fine-grained sediment deposition in continental freshwater lacustrine basins,with a focus on the Chang 7_(3) sub-member of Triassic Yanchang Formation.The research integrates a variety of exploration data,including field outcrops,drilling,logging,core samples,geochemical analyses,and flume simulation.The study indicates that:(1)The paleoenvironment of the Chang 7_(3) deposition is characterized by a warm and humid climate,frequent monsoon events,and a large water depth of freshwater lacustrine basin.The paleogeomorphology exhibits an asymmetrical pattern,with steep slopes in the southwest and gentle slopes in the northeast,which can be subdivided into microgeomorphological units,including depressions and ridges in lakebed,as well as ancient channels.(2)The Chang 7_(3) sub-member is characterized by a diverse array of fine-grained sediments,including very fine sandstone,siltstone,mudstone and tuff.These sediments are primarily distributed in thin interbedded and laminated arrangements vertically.The overall grain size of the sandstone predominantly falls below 62.5μm,with individual layer thicknesses of 0.05–0.64 m.The deposits contain intact plant fragments and display various sedimentary structure,such as wavy bedding,inverse-to-normal grading sequence,and climbing ripple bedding,which indicating a depositional origin associated with density flows.(3)Flume simulation experiments have successfully replicated the transport processes and sedimentary characteristics associated with density flows.The initial phase is characterized by a density-velocity differential,resulting in a thicker,coarser sediment layer at the flow front,while the upper layers are thinner and finer in grain size.During the mid-phase,sliding water effects cause the fluid front to rise and facilitate rapid forward transport.This process generates multiple“new fronts”,enabling the long-distance transport of fine-grained sandstones,such as siltstone and argillaceous siltstone,into the center of the lake basin.(4)A sedimentary model primarily controlled by hyperpynal flows was established for the southwestern part of the basin,highlighting that the frequent occurrence of flood events and the steep slope topography in this area are primary controlling factors for the development of hyperpynal flows.(5)Sandstone and mudstone in the Chang 7_(3) sub-member exhibit micro-and nano-scale pore-throat systems,shale oil is present in various lithologies,while the content of movable oil varies considerably,with sandstone exhibiting the highest content of movable oil.(6)The fine-grained sediment complexes formed by multiple episodes of sandstones and mudstones associated with density flow in the Chang 7_(3) formation exhibit characteristics of“overall oil-bearing with differential storage capacity”.The combination of mudstone with low total organic carbon content(TOC)and siltstone is identified as the most favorable exploration target at present.
基金Supported by the National Natural Science Foundation of China(61601176)。
文摘In this paper,we propose hierarchical attention dual network(DNet)for fine-grained image classification.The DNet can randomly select pairs of inputs from the dataset and compare the differences between them through hierarchical attention feature learning,which are used simultaneously to remove noise and retain salient features.In the loss function,it considers the losses of difference in paired images according to the intra-variance and inter-variance.In addition,we also collect the disaster scene dataset from remote sensing images and apply the proposed method to disaster scene classification,which contains complex scenes and multiple types of disasters.Compared to other methods,experimental results show that the DNet with hierarchical attention is robust to different datasets and performs better.
基金supported by the National Natural Science Foundation of China(Nos.11872216 and 52071139)the Natural Science Foundation of Hunan Province(Nos.2019JJ50586 and 2019JJ50591)the Scientific Research Project of Hunan Provincial Department of Education(Nos.18B193 and 21B0471).
文摘To understand the mechanical and twinning behaviors of a fine-grained extruded Mg-8.0Al-0.1Mn-2.0Ca alloy under high-speed impact,impact tests were carried out using a split Hopkinson pressure bar,and microstructures at strains of 0.05,0.10 and 0.20 were obtained using a series of stop rings manufactured by high-strength steel.The stress response and twinning behavior are closely related to loading direction and applied strain rate.The true stress-true strain curves are s-shaped in extrusion direction(ED)and c-shaped in transverse direction(TD),showing apparent anisotropy,while the yield strength is insensitive to loading direction.Almost identical strain-rate sensitivity is demonstrated by the stress in ED and TD.Interestingly,de-twinning is apparent as the applied strain increases to 0.20,and it is enhanced with increasing the applied strain rate.In contrast,the twin density in ED samples is clearly higher than that in TD samples.By modifying the terms of strain hardening and strain rate hardening in the classical JC model,an optimized model is built,which can accurately predict the stress response behavior of the studied alloy under high-speed impact along ED and TD.The correlation coefficient(R)and average absolute relative error(AARE)are 98.63%and 0.0199 for ED,and 96.88%and 0.0202 for TD,respectively.
基金The financial support of the National Natural Science Foundation of China(Grant Nos.41972293,42272337)the Science Fund for Distinguished Young Scholars of Hubei Province(Grant No.2023AFA078)are gratefully acknowledged.
文摘Soil responds to cavity expansion is inherently rate-dependent,especially in the case of fine-grained soils.To better understand such rate effects,self-boring pressuremeter tests were conducted on Kunming peaty soil within a strain rate range of 0.1%/min to 5.0%/min.The results showed a clear dependence of cavity pressure and excess pore pressure(EPP)on strain ratesdboth increased with higher rates for a given radial displacement.In light of the experimental results,three cases of cylindrical cavity expansion were investigated using the finite element method and analytical method,partially drained expansion in Modified Cam-Clay(MCC)soil,and undrained and partially drained expansion in elastoviscoplastic(EVP)soil.The EVP behavior was and modeled using the MCC model and the overstress viscoplastic theory.The results indicated that over the strain rate range of 0.0001%/min and 50%/min,the rate response of cavity pressure for the case of partially drained expansion in MCC soil(permeability coefficient ranging from 5×10^(-6) m/s to 2.5×10^(-11) m/s)is not obvious,while the EPP response during undrained expansion in EVP soil shows rate-independent.Only the partially drained solution for cavity expansion in EVP soil captured the rate-sensitive responses of both cavity pressure and EPP,confirmed by the pressuremeter tests on the Kunming peaty soil,Saint-Herblain clay,and Burswood clay.This suggests that the rate effect results from a combination of drainage-related and time-dependent soil behavior.Parametric studies further demonstrated that both viscous behavior and the overconsolidation ratio significantly influence cylindrical cavity expansion response,and the drainage conditions during expansion can be assessed using a nondimensional velocity.
基金supported by the National Natural Science Foundation of China(32171777)the Natural Science Foundation of Heilongjiang for Distinguished Young Scientists(JQ2023F002)the Fundamental Research Funds for Central Universities(2572023CT16).
文摘Climate change is a global phenomenon that has profound impacts on ecological dynamics and biodiversity,shaping the interactions between species and their environment.To gain a deeper understanding of the mechanisms driving climate change,phenological monitoring is essential.Traditional methods of defining phenological phases often rely on fixed thresholds.However,with the development of technology,deep learning-based classification models are now able to more accurately delineate phenological phases from images,enabling phenological monitoring.Despite the significant advancements these models have made in phenological monitoring,they still face challenges in fully capturing the complexity of biotic-environmental interactions,which can limit the fine-grained accuracy of phenological phase identification.To address this,we propose a novel deep learning model,RESformer,designed to monitor tree phenology at a fine-grained level using PhenoCam images.RESformer features a lightweight structure,making it suitable for deployment in resource-constrained environments.It incorporates a dual-branch routing mechanism that considers both global and local information,thereby improving the accuracy of phenological monitoring.To validate the effectiveness of RESformer,we conducted a case study involving 82,118 images taken over two years from four different locations in Wisconsin,focusing on the phenology of Acer.The images were classified into seven distinct phenological stages,with RESformer achieving an overall monitoring accuracy of 96.02%.Furthermore,we compared RESformer with a phenological monitoring approach based on the Green Chromatic Coordinate(GCC)index and ten popular classification models.The results showed that RESformer excelled in fine-grained monitoring,effectively capturing and identifying changes in phenological stages.This finding not only provides strong support for monitoring the phenology of Acer species but also offers valuable insights for understanding ecological trends and developing more effective ecosystem conservation and management strategies.
基金financial support by the National Natural Science Foundation of China(No.51364032)the Inner Mongolia Natural Science Foundation(No.2022MS05028)。
文摘The spray-deposition was used to produce billets of Mg-4Al-1.5Zn-3Ca-1Nd(A alloy)and Mg-13Al-3Zn-3Ca-1Nd(B alloy),and evolution of deformation substructure and Mg_(x)Zn_(y)Ca_(z)metastable phase in fine-grained(3μm)Mg alloys was investigated by scanning electron microscopy(SEM),transmission electron microscopy(TEM),X-ray diffraction(XRD),and electron backscattered diffraction(EBSD).It was found that different dislocation configurations were formed in A and B alloys.Redundant free dislocations(RFDs)and dislocation tangles were the ways to form deformation substructure in A alloy,no RFDs except dislocation tangles were found in B alloy.The interaction between nano-scale second phase particles(nano-scale C15 andβ-Mg_(17)(Al,Zn)_(12)phase)and different dislocation configurations had a significant effect on the deformation substructures formation.The mass transfer of Mg_(x)Zn_(y)Ca_(z)metastable phases and the stacking order of stacking faults were conducive to the Mg-Nd-Zn typed long period stacking ordered(LPSO)phases formation.Nano-scale C15 phases,Mg-Nd-Zn typed LPSO phases,c/a ratio,β-Mg_(17)(Al,Zn)_(12)phases were the key factors influencing the formation of textures.Different textures and grain boundary features(GB features)had a significant effect on k-value.The non-basal textures were the main factor affecting k-value in A alloy,while the high-angle grain boundary(HAGB)was the main factor affecting k-value in B alloy.
基金supported by the Beijing Natural Science Foundation(No.5252014)the Open Fund of The Key Laboratory of Urban Ecological Environment Simulation and Protection,Ministry of Ecology and Environment of the People's Republic of China (No.UEESP-202502)the National Natural Science Foundation of China (No.62303063&32371874)。
文摘Bird monitoring and protection are essential for maintaining biodiversity,and fine-grained bird classification has become a key focus in this field.Audio-visual modalities provide critical cues for this task,but robust feature extraction and efficient fusion remain major challenges.We introduce a multi-stage fine-grained audiovisual fusion network(MSFG-AVFNet) for fine-grained bird species classification,which addresses these challenges through two key components:(1) the audiovisual feature extraction module,which adopts a multi-stage finetuning strategy to provide high-quality unimodal features,laying a solid foundation for modality fusion;(2) the audiovisual feature fusion module,which combines a max pooling aggregation strategy with a novel audiovisual loss function to achieve effective and robust feature fusion.Experiments were conducted on the self-built AVB81and the publicly available SSW60 datasets,which contain data from 81 and 60 bird species,respectively.Comprehensive experiments demonstrate that our approach achieves notable performance gains,outperforming existing state-of-the-art methods.These results highlight its effectiveness in leveraging audiovisual modalities for fine-grained bird classification and its potential to support ecological monitoring and biodiversity research.
基金supported by National Natural Science Foundation of China(No.62471034)Hebei Natural Science Foundation(No.F2023105001).
文摘Fine-grained aircraft target detection in remote sensing holds significant research valueand practical applications,particularly in military defense and precision strikes.Given the complex-ity of remote sensing images,where targets are often small and similar within categories,detectingthese fine-grained targets is challenging.To address this,we constructed a fine-grained dataset ofremotely sensed airplanes;for the problems of remote sensing fine-grained targets with obvious head-to-tail distributions and large variations in target sizes,we proposed the DWDet fine-grained tar-get detection and recognition algorithm.First,for the problem of unbalanced category distribution,we adopt an adaptive sampling strategy.In addition,we construct a deformable convolutional blockand improve the decoupling head structure to improve the detection effect of the model ondeformed targets.Then,we design a localization loss function,which is used to improve the model’slocalization ability for targets of different scales.The experimental results show that our algorithmimproves the overall accuracy of the model by 4.1%compared to the baseline model,and improvesthe detection accuracy of small targets by 12.2%.The ablation and comparison experiments alsoprove the effectiveness of our algorithm.
基金supported by the National Natural Science Foundation of China,China (Grants No.62171232)the Priority Academic Program Development of Jiangsu Higher Education Institutions,China。
文摘Fine-grained Image Recognition(FGIR)task is dedicated to distinguishing similar sub-categories that belong to the same super-category,such as bird species and car types.In order to highlight visual differences,existing FGIR works often follow two steps:discriminative sub-region localization and local feature representation.However,these works pay less attention on global context information.They neglect a fact that the subtle visual difference in challenging scenarios can be highlighted through exploiting the spatial relationship among different subregions from a global view point.Therefore,in this paper,we consider both global and local information for FGIR,and propose a collaborative teacher-student strategy to reinforce and unity the two types of information.Our framework is implemented mainly by convolutional neural network,referred to Teacher-Student Based Attention Convolutional Neural Network(T-S-ACNN).For fine-grained local information,we choose the classic Multi-Attention Network(MA-Net)as our baseline,and propose a type of boundary constraint to further reduce background noises in the local attention maps.In this way,the discriminative sub-regions tend to appear in the area occupied by fine-grained objects,leading to more accurate sub-region localization.For fine-grained global information,we design a graph convolution based Global Attention Network(GA-Net),which can combine extracted local attention maps from MA-Net with non-local techniques to explore spatial relationship among subregions.At last,we develop a collaborative teacher-student strategy to adaptively determine the attended roles and optimization modes,so as to enhance the cooperative reinforcement of MA-Net and GA-Net.Extensive experiments on CUB-200-2011,Stanford Cars and FGVC Aircraft datasets illustrate the promising performance of our framework.
基金supported by the Fundamental Research Funds for the Central Universities(Grant No.2022CDJQY-012)the Innovation Group Science Foundation of the Natural Science Foundation of Chongqing,China(Grant No.cstc2020jcyj-cxttX0003).
文摘Coralline soils,specialized materials found extensively in the South China Sea,are playing an increasingly vital role in engineering projects.However,like most terrigenous soils,fine-grained coral soil is prone to shrinkage and cracking,which can significantly affect its engineering properties and ultimately jeopardize engineering safety.This paper presents a desiccation cracking test of fine-grained coral soil,with a particular focus on the thickness effect.The study involved measuring the water content and recording the evolution of desiccation cracking.Advanced image processing technology is employed to analyze the variations in crack parameters,clod parameters,fractal dimensions,frequency distributions,and desiccation cracking propagation velocities of fine-grained coral soil.Furthermore,the dynamic evolution of desiccation cracking under the influence of layer thickness is analyzed.A comprehensive crack evolution model is proposed,encompassing both top-down and bottom-up crack propagation,as well as internal tensile cracking.This work introduces novel metrics for the propagation velocity of the total crack area,the characteristic propagation velocities of desiccation cracks,and the acceleration of crack propagation.Through data fitting,theoretical formulas for soil water evaporation,propagation velocities of desiccation cracks,and crack propagation acceleration are derived,laying a foundation for future soil cracking theories.
基金supported by the BK21 FOUR Program of the National Research Foundation of Korea funded by the Ministry of Education(NRF5199991014091)Seok-Won Lee’s work was supported by Institute of Information&Communications Technology Planning&Evaluation(IITP)under the Artificial Intelligence Convergence Innovation Human Resources Development(IITP-2024-RS-2023-00255968)grant funded by the Korea government(MSIT).
文摘Although sentiment analysis is pivotal to understanding user preferences,existing models face significant challenges in handling context-dependent sentiments,sarcasm,and nuanced emotions.This study addresses these challenges by integrating ontology-based methods with deep learning models,thereby enhancing sentiment analysis accuracy in complex domains such as film reviews and restaurant feedback.The framework comprises explicit topic recognition,followed by implicit topic identification to mitigate topic interference in subsequent sentiment analysis.In the context of sentiment analysis,we develop an expanded sentiment lexicon based on domainspecific corpora by leveraging techniques such as word-frequency analysis and word embedding.Furthermore,we introduce a sentiment recognition method based on both ontology-derived sentiment features and sentiment lexicons.We evaluate the performance of our system using a dataset of 10,500 restaurant reviews,focusing on sentiment classification accuracy.The incorporation of specialized lexicons and ontology structures enables the framework to discern subtle sentiment variations and context-specific expressions,thereby improving the overall sentiment-analysis performance.Experimental results demonstrate that the integration of ontology-based methods and deep learning models significantly improves sentiment analysis accuracy.
基金the Open Project of Sichuan Provincial Key Laboratory of Philosophy and Social Science for Language Intelligence in Special Education under Grant No.YYZN-2023-4the Ph.D.Fund of Chengdu Technological University under Grant No.2020RC002.
文摘The fingerprinting-based approach using the wireless local area network(WLAN)is widely used for indoor localization.However,the construction of the fingerprint database is quite time-consuming.Especially when the position of the access point(AP)or wall changes,updating the fingerprint database in real-time is difficult.An appropriate indoor localization approach,which has a low implementation cost,excellent real-time performance,and high localization accuracy and fully considers complex indoor environment factors,is preferred in location-based services(LBSs)applications.In this paper,we proposed a fine-grained grid computing(FGGC)model to achieve decimeter-level localization accuracy.Reference points(RPs)are generated in the grid by the FGGC model.Then,the received signal strength(RSS)values at each RP are calculated with the attenuation factors,such as the frequency band,three-dimensional propagation distance,and walls in complex environments.As a result,the fingerprint database can be established automatically without manual measurement,and the efficiency and cost that the FGGC model takes for the fingerprint database are superior to previous methods.The proposed indoor localization approach,which estimates the position step by step from the approximate grid location to the fine-grained location,can achieve higher real-time performance and localization accuracy simultaneously.The mean error of the proposed model is 0.36 m,far lower than that of previous approaches.Thus,the proposed model is feasible to improve the efficiency and accuracy of Wi-Fi indoor localization.It also shows high-accuracy performance with a fast running speed even under a large-size grid.The results indicate that the proposed method can also be suitable for precise marketing,indoor navigation,and emergency rescue.
基金Supported by the National Science and Technology Major Project of China(2017ZX05009-002)National Natural Science Foundation of China(41772090)。
文摘Based on sedimentary characteristics of the fine-grained rocks of the lower submember of second member of the Lower Cretaceous Shahezi Formation(K_(1)sh_(2)^(L))in the Lishu rift depression,combined with methods of organic petrology,analysis of major and trace elements as well as biological marker compound,the enrichment conditions and enrichment model of organic matter in the fine-grained sedimentary rocks in volcanic rift lacustrine basin are investigated.The change of sedimentary paleoenvironment controls the vertical distribution of different lithofacies types in the K_(1)sh_(2)^(L)and divides it into the upper and lower parts.The lower part contains massive siliceous mudstone with bioclast-bearing siliceous mudstone,whereas the upper part is mostly composed of laminated siliceous shale and laminated fine-grained mixed shale.The kerogen types of organic matter in the lower and upper parts are typesⅡ_(2)–Ⅲand typesⅠ–Ⅱ_(1),respectively.The organic carbon content in the upper part is higher than that in the lower part generally.The enrichment of organic matter in volcanic rift lacustrine basin is subjected to three favorable conditions.First,continuous enhancement of rifting is the direct factor increasing the paleo-water depth,and the rise of base level leads to the expansion of deep-water mudstone/shale deposition range.Second,relatively strong underwater volcanic eruption and rifting are simultaneous,and such event can provide a lot of nutrients for the lake basin,which is conducive to the bloom of algae,resulting in higher productivity of typesⅠ–Ⅱ_(1)kerogen.Third,the relatively dry paleoclimate leads to a decrease in input of fresh water and terrestrial materials,including TypeⅢkerogen from terrestrial higher plants,resulting in a water body with higher salinity and anoxic stratification,which is more favorable for preservation of organic matter.The organic matter enrichment model of fine-grained sedimentary rocks of volcanic rift lacustrine basin is established,which is of reference significance to the understanding of the organic matter enrichment mechanism of fine-grained sedimentary rocks of Shahezi Formation in Songliao Basin and even in the northeast China.
基金supported by the Fundamental Research Funds for the Central Universities(2024JKF13)the Beijing Municipal Education Commission General Program of Science and Technology(No.KM202414019003).
文摘With the widespread use of SMS(Short Message Service),the proliferation of malicious SMS has emerged as a pressing societal issue.While deep learning-based text classifiers offer promise,they often exhibit suboptimal performance in fine-grained detection tasks,primarily due to imbalanced datasets and insufficient model representation capabilities.To address this challenge,this paper proposes an LLMs-enhanced graph fusion dual-stream Transformer model for fine-grained Chinese malicious SMS detection.During the data processing stage,Large Language Models(LLMs)are employed for data augmentation,mitigating dataset imbalance.In the data input stage,both word-level and character-level features are utilized as model inputs,enhancing the richness of features and preventing information loss.A dual-stream Transformer serves as the backbone network in the learning representation stage,complemented by a graph-based feature fusion mechanism.At the output stage,both supervised classification cross-entropy loss and supervised contrastive learning loss are used as multi-task optimization objectives,further enhancing the model’s feature representation.Experimental results demonstrate that the proposed method significantly outperforms baselines on a publicly available Chinese malicious SMS dataset.
