It is of great significance to study the failure mode of mining roadways for safe coal mining.The unconventional asymmetric failure(UAF)phenomenon was discovered in the 9106 ventilation roadway of Wangzhuang coal mine...It is of great significance to study the failure mode of mining roadways for safe coal mining.The unconventional asymmetric failure(UAF)phenomenon was discovered in the 9106 ventilation roadway of Wangzhuang coal mine in Shanxi Province.The main manifestation is that the deformation of the roadway on the coal side is much greater than that on the coal pillar side.A comprehensive study was conducted on on-site detection,theoretical analysis,laboratory tests and numerical simulation of the UAF phenomenon.On-site detection shows that the deformation of the coal sidewall can reach 50–80 cm,and the failure zone depth can reach 3 m.The deformation and fracture depth on the coal pillar side are much smaller than those on the coal side.A calculation model for the principal stress of surrounding rock when the axial direction of the roadway is inconsistent with the in-situ stress field was established.The distribution of the failure zone on both sides of the roadway has been defined by the combined mining induced stress.The true triaxial test studied the mechanical mechanism of rock mass fracture and crack propagation on both sides of the roadway.The research results indicate that the axial direction,stress field distribution,and mining induced stress field distribution of the roadway jointly affect the asymmetric failure mode of the roadway.The angle between the axis direction of the roadway and the maximum horizontal stress field leads to uneven distribution of the principal stress field on both sides.The differential distribution of mining induced stress exacerbates the asymmetric distribution of principal stress in the surrounding rock.The uneven stress distribution on both sides of the roadway is the main cause of UAF formation.The research results can provide mechanical explanations and theoretical support for the control of surrounding rock in roadways with similar failure characteristics.展开更多
Tomato is a major economic crop worldwide,and diseases on tomato leaves can significantly reduce both yield and quality.Traditional manual inspection is inefficient and highly subjective,making it difficult to meet th...Tomato is a major economic crop worldwide,and diseases on tomato leaves can significantly reduce both yield and quality.Traditional manual inspection is inefficient and highly subjective,making it difficult to meet the requirements of early disease identification in complex natural environments.To address this issue,this study proposes an improved YOLO11-based model,YOLO-SPDNet(Scale Sequence Fusion,Position-Channel Attention,and Dual Enhancement Network).The model integrates the SEAM(Self-Ensembling Attention Mechanism)semantic enhancement module,the MLCA(Mixed Local Channel Attention)lightweight attention mechanism,and the SPA(Scale-Position-Detail Awareness)module composed of SSFF(Scale Sequence Feature Fusion),TFE(Triple Feature Encoding),and CPAM(Channel and Position Attention Mechanism).These enhancements strengthen fine-grained lesion detection while maintaining model lightweightness.Experimental results show that YOLO-SPDNet achieves an accuracy of 91.8%,a recall of 86.5%,and an mAP@0.5 of 90.6%on the test set,with a computational complexity of 12.5 GFLOPs.Furthermore,the model reaches a real-time inference speed of 987 FPS,making it suitable for deployment on mobile agricultural terminals and online monitoring systems.Comparative analysis and ablation studies further validate the reliability and practical applicability of the proposed model in complex natural scenes.展开更多
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
Image steganography is a technique that hides secret information into the cover image to protect information security.The current image steganography is mainly to embed a smaller secret image in an area such as a text...Image steganography is a technique that hides secret information into the cover image to protect information security.The current image steganography is mainly to embed a smaller secret image in an area such as a texture of a larger-sized cover image,which will cause the size of the secret image to be much smaller than the cover image.Therefore,the problem of small steganographic capacity needs to be solved urgently.This paper proposes a steganography framework that combines image compression.In this framework,the Vector Quantized Variational AutoEncoder(VQ-VAE)is used to achieve the compression of the secret image.The compressed and reconstructed image is visually indistinguishable from the original image and facilitates more embedded data information later.Finally,the compressed image is transmitted to a SegNet deep neural network that contains a set of encoders and decoders to achieve image hiding and extraction.Experimental results show that the steganographic framework guarantees the quality of steganography while its relative steganographic capacity reaches 1.Besides,Peak Signal-to-Noise Ratio(PSNR)and Structural Similarity Index(SSIM)values can reach 42 dB and 0.94,respectively.展开更多
Mobile bike-sharing services have been prevalently used in many cities as an important urban commuting service and a promising way to build smart cities,especially in the new era of 5G and Internet-of-Things(IoT)envir...Mobile bike-sharing services have been prevalently used in many cities as an important urban commuting service and a promising way to build smart cities,especially in the new era of 5G and Internet-of-Things(IoT)environments.A mobile bike-sharing service makes commuting convenient for people and imparts new vitality to urban transportation systems.In the real world,the problems of no docks or no bikes at bike-sharing stations often arise because of several inevitable reasons such as the uncertainty of bike usage.In addition to pure manual rebalancing,in several works,attempts were made to predict the demand for bikes.In this paper,we devised a bike-sharing service with highly accurate demand prediction using collaborative computing and information fusion.We combined the information of bike demands at different time periods and the locations between stations and proposed a dynamical clustering algorithm for station clustering.We carefully analyzed and discovered the group of features that impact the demand of bikes,from historical bike-sharing records and 5G IoT environment data.We combined the discovered information and proposed an XGBoost-based regression model to predict the rental and return demand.We performed sufficient experiments on two real-world datasets.The results confirm that compared to some existing methods,our method produces superior prediction results and performance and improves the availability of bike-sharing service in 5G IoT environments.展开更多
This paper presents a new idea, named as modeling multisensor-heterogeneous information, to incorporate the fuzzy logic methodologies with mulitsensor-multitarget system under the framework of random set theory. First...This paper presents a new idea, named as modeling multisensor-heterogeneous information, to incorporate the fuzzy logic methodologies with mulitsensor-multitarget system under the framework of random set theory. Firstly, based on strong random set and weak random set, the unified form to describe both data (unambiguous information) and fuzzy evidence (uncertain information) is introduced. Secondly, according to signatures of fuzzy evidence, two Bayesian-markov nonlinear measurement models are proposed to fuse effectively data and fuzzy evidence. Thirdly, by use of "the models-based signature-matching scheme", the operation of the statistics of fuzzy evidence defined as random set can be translated into that of the membership functions of relative point state variables. These works are the basis to construct qualitative measurement models and to fuse data and fuzzy evidence.展开更多
The modifiedλ-differential Lie-Yamaguti algebras are considered,in which a modifiedλ-differential Lie-Yamaguti algebra consisting of a Lie-Yamaguti algebra and a modifiedλ-differential operator.First we introduce t...The modifiedλ-differential Lie-Yamaguti algebras are considered,in which a modifiedλ-differential Lie-Yamaguti algebra consisting of a Lie-Yamaguti algebra and a modifiedλ-differential operator.First we introduce the representation of modifiedλ-differential Lie-Yamaguti algebras.Furthermore,we establish the cohomology of a modifiedλ-differential Lie-Yamaguti algebra with coefficients in a representation.Finally,we investigate the one-parameter formal deformations and Abelian extensions of modifiedλ-differential Lie-Yamaguti algebras using the second cohomology group.展开更多
This paper proposed an efficient method of image overlapping relationship analysis based on spatial index of KD tree fast search for disordered and large-scale asteroid images.In this study,the image data from asteroi...This paper proposed an efficient method of image overlapping relationship analysis based on spatial index of KD tree fast search for disordered and large-scale asteroid images.In this study,the image data from asteroid exploration missions such as Bennu,Vesta,and Ryugu were used for experiments,and the proposed image matching pairs determination algorithm was comprehensively compared with the corresponding modules of USGS ISIS in order to evaluate its performance in terms of efficiency and accuracy.The results show that when processing more than a thousand images,the proposed method greatly improves the efficiency of acquiring image matching pairs while ensuring the correctness of image overlapping relationships and accuracy of bundle adjustment.At the same time,according to the obtained image matching pairs,images that meet the requirements of Stereo Photoclinometry can be quickly selected,effectively improving the quality of 3D reconstruction models of asteroid images.展开更多
This paper aims to enhance the array Beamforming(BF) robustness by tackling issues related to BF weight state estimation encountered in Constant Modulus Blind Beamforming(CMBB). To achieve this, we introduce a novel a...This paper aims to enhance the array Beamforming(BF) robustness by tackling issues related to BF weight state estimation encountered in Constant Modulus Blind Beamforming(CMBB). To achieve this, we introduce a novel approach that incorporates an L1-regularizer term in BF weight state estimation. We start by explaining the CMBB formation mechanism under conditions where there is a mismatch in the far-field signal model. Subsequently, we reformulate the BF weight state estimation challenge using a method known as variable-splitting, turning it into a noise minimization problem. This problem combines both linear and nonlinear quadratic terms with an L1-regularizer that promotes the sparsity. The optimization strategy is based on a variable-splitting method, implemented using the Alternating Direction Method of Multipliers(ADMM). Furthermore, a variable-splitting framework is developed to enhance BF weight state estimation, employing a Kalman Smoother(KS) optimization algorithm. The approach integrates the Rauch-TungStriebel smoother to perform posterior-smoothing state estimation by leveraging prior data. We provide proof of convergence for both linear and nonlinear CMBB state estimation technology using the variable-splitting KS and the iterated extended Kalman smoother. Simulations corroborate our theoretical analysis, showing that the proposed method achieves robust stability and effective convergence, even when faced with signal model mismatches.展开更多
Er^(3+)-doped BaLaGaO_(4)green phosphors was synthesized through a high-temperature solid-state reaction technique.The phase structure and morphology test results of the phosphor indicate that the BaLaGaO_(4)material ...Er^(3+)-doped BaLaGaO_(4)green phosphors was synthesized through a high-temperature solid-state reaction technique.The phase structure and morphology test results of the phosphor indicate that the BaLaGaO_(4)material was successfully synthesized and Er^(3+)ions were successfully doped into the main lattice.This doping does change the basic structure of the crystal.BaLaGaO_(4):Er^(3+)phosphor exhibits bright green emission centered at 545 nm when excited by 381 nm ultraviolet light or 980 nm near-infrared light.The optimal doping concentration is found to be x=0.04.To quantify the temperature sensitivity of the phosphor,the fluorescence intensity ratio method was used.Within the temperature range of 298-473 K,the maximum relative sensitivities are 1.35%/K(298 K,381 nm)and 1.45%/K(298 K,980 nm),respectively.The maximum absolute sensitivities are 0.67%/K(473 K,381 nm)and 0.69%/K(473 K,980 nm),respectively.Finally,white light-emitting diodes(WLEDs)with a high colour index of Ra=82and a relatively low correlated colour temperature of CCT=5064 K are obtained by integrating the synthesized BaLaGaO_(4):0.04Er^(3+)green phosphor into warm WLEDs devices.These results suggest that Er^(3+)-activated BaLaGaO_(4)multifunctional phosphors hold considerable promise in the areas of optical temperature sensing and WLEDs phosphor conversion.展开更多
Deep neural network(DNN)models have achieved remarkable performance across diverse tasks,leading to widespread commercial adoption.However,training high-accuracy models demands extensive data,substantial computational...Deep neural network(DNN)models have achieved remarkable performance across diverse tasks,leading to widespread commercial adoption.However,training high-accuracy models demands extensive data,substantial computational resources,and significant time investment,making them valuable assets vulnerable to unauthorized exploitation.To address this issue,this paper proposes an intellectual property(IP)protection framework for DNN models based on feature layer selection and hyper-chaotic mapping.Firstly,a sensitivity-based importance evaluation algorithm is used to identify the key feature layers for encryption,effectively protecting the core components of the model.Next,the L1 regularization criterion is applied to further select high-weight features that significantly impact the model’s performance,ensuring that the encryption process minimizes performance loss.Finally,a dual-layer encryption mechanism is designed,introducing perturbations into the weight values and utilizing hyperchaotic mapping to disrupt channel information,further enhancing the model’s security.Experimental results demonstrate that encrypting only a small subset of parameters effectively reduces model accuracy to random-guessing levels while ensuring full recoverability.The scheme exhibits strong robustness against model pruning and fine-tuning attacks and maintains consistent performance across multiple datasets,providing an efficient and practical solution for authorization-based DNN IP protection.展开更多
Lunar wrinkle ridges are an important stress geological structure on the Moon, which reflect the stress state and geological activity on the Moon. They provide important insights into the evolution of the Moon and are...Lunar wrinkle ridges are an important stress geological structure on the Moon, which reflect the stress state and geological activity on the Moon. They provide important insights into the evolution of the Moon and are key factors influencing future lunar activity, such as the choice of landing sites. However, automatic extraction of lunar wrinkle ridges is a challenging task due to their complex morphology and ambiguous features. Traditional manual extraction methods are time-consuming and labor-intensive. To achieve automated and detailed detection of lunar wrinkle ridges, we have constructed a lunar wrinkle ridge data set, incorporating previously unused aspect data to provide edge information, and proposed a Dual-Branch Ridge Detection Network(DBR-Net) based on deep learning technology. This method employs a dual-branch architecture and an Attention Complementary Feature Fusion module to address the issue of insufficient lunar wrinkle ridge features. Through comparisons with the results of various deep learning approaches, it is demonstrated that the proposed method exhibits superior detection performance. Furthermore, the trained model was applied to lunar mare regions, generating a distribution map of lunar mare wrinkle ridges;a significant linear relationship between the length and area of the lunar wrinkle ridges was obtained through statistical analysis, and six previously unrecorded potential lunar wrinkle ridges were detected. The proposed method upgrades the automated extraction of lunar wrinkle ridges to a pixel-level precision and verifies the effectiveness of DBR-Net in lunar wrinkle ridge detection.展开更多
In recent years,with the rapid development of software systems,the continuous expansion of software scale and the increasing complexity of systems have led to the emergence of a growing number of software metrics.Defe...In recent years,with the rapid development of software systems,the continuous expansion of software scale and the increasing complexity of systems have led to the emergence of a growing number of software metrics.Defect prediction methods based on software metric elements highly rely on software metric data.However,redundant software metric data is not conducive to efficient defect prediction,posing severe challenges to current software defect prediction tasks.To address these issues,this paper focuses on the rational clustering of software metric data.Firstly,multiple software projects are evaluated to determine the preset number of clusters for software metrics,and various clustering methods are employed to cluster the metric elements.Subsequently,a co-occurrence matrix is designed to comprehensively quantify the number of times that metrics appear in the same category.Based on the comprehensive results,the software metric data are divided into two semantic views containing different metrics,thereby analyzing the semantic information behind the software metrics.On this basis,this paper also conducts an in-depth analysis of the impact of different semantic view of metrics on defect prediction results,as well as the performance of various classification models under these semantic views.Experiments show that the joint use of the two semantic views can significantly improve the performance of models in software defect prediction,providing a new understanding and approach at the semantic view level for defect prediction research based on software metrics.展开更多
Magnetic resonance imaging(MRI)plays an important role in medical diagnosis,generating petabytes of image data annually in large hospitals.This voluminous data stream requires a significant amount of network bandwidth...Magnetic resonance imaging(MRI)plays an important role in medical diagnosis,generating petabytes of image data annually in large hospitals.This voluminous data stream requires a significant amount of network bandwidth and extensive storage infrastructure.Additionally,local data processing demands substantial manpower and hardware investments.Data isolation across different healthcare institutions hinders crossinstitutional collaboration in clinics and research.In this work,we anticipate an innovative MRI system and its four generations that integrate emerging distributed cloud computing,6G bandwidth,edge computing,federated learning,and blockchain technology.This system is called Cloud-MRI,aiming at solving the problems of MRI data storage security,transmission speed,artificial intelligence(AI)algorithm maintenance,hardware upgrading,and collaborative work.The workflow commences with the transformation of k-space raw data into the standardized Imaging Society for Magnetic Resonance in Medicine Raw Data(ISMRMRD)format.Then,the data are uploaded to the cloud or edge nodes for fast image reconstruction,neural network training,and automatic analysis.Then,the outcomes are seamlessly transmitted to clinics or research institutes for diagnosis and other services.The Cloud-MRI system will save the raw imaging data,reduce the risk of data loss,facilitate inter-institutional medical collaboration,and finally improve diagnostic accuracy and work efficiency.展开更多
With the advancements of the next-generation communication networking and Internet ofThings(IoT)technologies,a variety of computation-intensive applications(e.g.,autonomous driving and face recognition)have emerged.Th...With the advancements of the next-generation communication networking and Internet ofThings(IoT)technologies,a variety of computation-intensive applications(e.g.,autonomous driving and face recognition)have emerged.The execution of these IoT applications demands a lot of computing resources.Nevertheless,terminal devices(TDs)usually do not have sufficient computing resources to process these applications.Offloading IoT applications to be processed by mobile edge computing(MEC)servers with more computing resources provides a promising way to address this issue.While a significant number of works have studied task offloading,only a few of them have considered the security issue.This study investigates the problem of spectrum allocation and security-sensitive task offloading in an MEC system.Dynamic voltage scaling(DVS)technology is applied by TDs to reduce energy consumption and computing time.To guarantee data security during task offloading,we use AES cryptographic technique.The studied problem is formulated as an optimization problem and solved by our proposed efficient offloading scheme.The simulation results show that the proposed scheme can reduce system cost while guaranteeing data security.展开更多
Vehicle Edge Computing(VEC)and Cloud Computing(CC)significantly enhance the processing efficiency of delay-sensitive and computation-intensive applications by offloading compute-intensive tasks from resource-constrain...Vehicle Edge Computing(VEC)and Cloud Computing(CC)significantly enhance the processing efficiency of delay-sensitive and computation-intensive applications by offloading compute-intensive tasks from resource-constrained onboard devices to nearby Roadside Unit(RSU),thereby achieving lower delay and energy consumption.However,due to the limited storage capacity and energy budget of RSUs,it is challenging to meet the demands of the highly dynamic Internet of Vehicles(IoV)environment.Therefore,determining reasonable service caching and computation offloading strategies is crucial.To address this,this paper proposes a joint service caching scheme for cloud-edge collaborative IoV computation offloading.By modeling the dynamic optimization problem using Markov Decision Processes(MDP),the scheme jointly optimizes task delay,energy consumption,load balancing,and privacy entropy to achieve better quality of service.Additionally,a dynamic adaptive multi-objective deep reinforcement learning algorithm is proposed.Each Double Deep Q-Network(DDQN)agent obtains rewards for different objectives based on distinct reward functions and dynamically updates the objective weights by learning the value changes between objectives using Radial Basis Function Networks(RBFN),thereby efficiently approximating the Pareto-optimal decisions for multiple objectives.Extensive experiments demonstrate that the proposed algorithm can better coordinate the three-tier computing resources of cloud,edge,and vehicles.Compared to existing algorithms,the proposed method reduces task delay and energy consumption by 10.64%and 5.1%,respectively.展开更多
At inference time,deep neural networks are susceptible to backdoor attacks,which can produce attackercontrolled outputs when inputs contain carefully crafted triggers.Existing defense methods often focus on specific a...At inference time,deep neural networks are susceptible to backdoor attacks,which can produce attackercontrolled outputs when inputs contain carefully crafted triggers.Existing defense methods often focus on specific attack types or incur high costs,such as data cleaning or model fine-tuning.In contrast,we argue that it is possible to achieve effective and generalizable defense without removing triggers or incurring high model-cleaning costs.Fromthe attacker’s perspective and based on characteristics of vulnerable neuron activation anomalies,we propose an Adaptive Feature Injection(AFI)method for black-box backdoor detection.AFI employs a pre-trained image encoder to extract multi-level deep features and constructs a dynamic weight fusionmechanism for precise identification and interception of poisoned samples.Specifically,we select the control samples with the largest feature differences fromthe clean dataset via feature-space analysis,and generate blended sample pairs with the test sample using dynamic linear interpolation.The detection statistic is computed by measuring the divergence G(x)in model output responses.We systematically evaluate the effectiveness of AFI against representative backdoor attacks,including BadNets,Blend,WaNet,and IAB,on three benchmark datasets:MNIST,CIFAR-10,and ImageNet.Experimental results show that AFI can effectively detect poisoned samples,achieving average detection rates of 95.20%,94.15%,and 86.49%on these datasets,respectively.Compared with existing methods,AFI demonstrates strong cross-domain generalization ability and robustness to unknown attacks.展开更多
Multirotor has been applied to many military and civilian mission scenarios. From the perspective of reliability, it is difficult to ensure that multirotors do not generate hardware and software failures or performanc...Multirotor has been applied to many military and civilian mission scenarios. From the perspective of reliability, it is difficult to ensure that multirotors do not generate hardware and software failures or performance anomalies during the flight process. These failures and anomalies may result in mission interruptions, crashes, and even threats to the lives and property of human beings.Thus, the study of flight reliability problems of multirotors is conductive to the development of the drone industry and has theoretical significance and engineering value. This paper proposes a reliable flight performance assessment method of multirotors based on an Interacting Multiple Model Particle Filter(IMMPF) algorithm and health degree as the performance indicator. First, the multirotor is modeled by the Stochastic Hybrid System(SHS) model, and the problem of reliable flight performance assessment is formulated. In order to solve the problem, the IMMPF algorithm is presented to estimate the real-time probability distribution of hybrid state of the established SHS-based multirotor model, since it can decrease estimation errors compared with the standard interacting multiple model algorithm based on extended Kalman filter. Then, the reliable flight performance is assessed with health degree based on the estimation result. Finally, a case study of a multirotor suffering from sensor anomalies is presented to validate the effectiveness of the proposed method.展开更多
Radio frequency interference（RFI） is becoming more and more frequently, which makes it an important issue in SAR imaging.RFI presented in synthetic aperture radar either on purpose or inadvertent will distort the us...Radio frequency interference（RFI） is becoming more and more frequently, which makes it an important issue in SAR imaging.RFI presented in synthetic aperture radar either on purpose or inadvertent will distort the useful SAR echoes, thus degrade the SAR image quality.To resolve this issue, a long time study was carried out to study the characteristic of the RFI through the RFIaffected spaceborne and airborne SAR data.Based on the narrow band nature of RFI, this paper proposes a new process which contains both RFI detection and RFI suppression.A useful subband spectral kurtosis detector is first used to detect RFI, and then its results are used for RFI suppression.The proposed process has two advantages： one is the economization on the compute time for unnecessary interference suppression when no RFI existed; the other is improving the performance of the suppression method with knowing the exact position where RFI is.Moreover, the previous RFI suppression method––subband spectral cancelation（SSC） is supplemented and perfected.The subband division step is also elaborated detail in this paper.The experiment results show that the subband spectral kurtosis detector exhibits good performance in recognizing both weak and narrow-band RFI.In addition, the validity of the SSC method with subband spectral kurtosis detector is also validated on the real SAR echoes.展开更多
基金financially supported by the National Natural Science Foundation of China(Nos.52225404,12532020,52394192 and 42321002)Key Research and Development Program Projects of Xinjiang Uygur Autonomous Region(No.2024B03017)Doctoral Startup Foundation of Fuyang Normal University,China(No.2025KYQD0124)。
文摘It is of great significance to study the failure mode of mining roadways for safe coal mining.The unconventional asymmetric failure(UAF)phenomenon was discovered in the 9106 ventilation roadway of Wangzhuang coal mine in Shanxi Province.The main manifestation is that the deformation of the roadway on the coal side is much greater than that on the coal pillar side.A comprehensive study was conducted on on-site detection,theoretical analysis,laboratory tests and numerical simulation of the UAF phenomenon.On-site detection shows that the deformation of the coal sidewall can reach 50–80 cm,and the failure zone depth can reach 3 m.The deformation and fracture depth on the coal pillar side are much smaller than those on the coal side.A calculation model for the principal stress of surrounding rock when the axial direction of the roadway is inconsistent with the in-situ stress field was established.The distribution of the failure zone on both sides of the roadway has been defined by the combined mining induced stress.The true triaxial test studied the mechanical mechanism of rock mass fracture and crack propagation on both sides of the roadway.The research results indicate that the axial direction,stress field distribution,and mining induced stress field distribution of the roadway jointly affect the asymmetric failure mode of the roadway.The angle between the axis direction of the roadway and the maximum horizontal stress field leads to uneven distribution of the principal stress field on both sides.The differential distribution of mining induced stress exacerbates the asymmetric distribution of principal stress in the surrounding rock.The uneven stress distribution on both sides of the roadway is the main cause of UAF formation.The research results can provide mechanical explanations and theoretical support for the control of surrounding rock in roadways with similar failure characteristics.
基金Tianmin Tianyuan Boutique Vegetable Industry Technology Service Station(Grant No.2024120011003081)Development of Environmental Monitoring and Traceability System for Wuqing Agricultural Production Areas(Grant No.2024120011001866)。
文摘Tomato is a major economic crop worldwide,and diseases on tomato leaves can significantly reduce both yield and quality.Traditional manual inspection is inefficient and highly subjective,making it difficult to meet the requirements of early disease identification in complex natural environments.To address this issue,this study proposes an improved YOLO11-based model,YOLO-SPDNet(Scale Sequence Fusion,Position-Channel Attention,and Dual Enhancement Network).The model integrates the SEAM(Self-Ensembling Attention Mechanism)semantic enhancement module,the MLCA(Mixed Local Channel Attention)lightweight attention mechanism,and the SPA(Scale-Position-Detail Awareness)module composed of SSFF(Scale Sequence Feature Fusion),TFE(Triple Feature Encoding),and CPAM(Channel and Position Attention Mechanism).These enhancements strengthen fine-grained lesion detection while maintaining model lightweightness.Experimental results show that YOLO-SPDNet achieves an accuracy of 91.8%,a recall of 86.5%,and an mAP@0.5 of 90.6%on the test set,with a computational complexity of 12.5 GFLOPs.Furthermore,the model reaches a real-time inference speed of 987 FPS,making it suitable for deployment on mobile agricultural terminals and online monitoring systems.Comparative analysis and ablation studies further validate the reliability and practical applicability of the proposed model in complex natural scenes.
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The paper was supported by the National Natural Science Foundation of China(61672354)the key scientific research project of Henan Provincial Higher Education(Nos.19B510005 and 20B413004).
文摘Image steganography is a technique that hides secret information into the cover image to protect information security.The current image steganography is mainly to embed a smaller secret image in an area such as a texture of a larger-sized cover image,which will cause the size of the secret image to be much smaller than the cover image.Therefore,the problem of small steganographic capacity needs to be solved urgently.This paper proposes a steganography framework that combines image compression.In this framework,the Vector Quantized Variational AutoEncoder(VQ-VAE)is used to achieve the compression of the secret image.The compressed and reconstructed image is visually indistinguishable from the original image and facilitates more embedded data information later.Finally,the compressed image is transmitted to a SegNet deep neural network that contains a set of encoders and decoders to achieve image hiding and extraction.Experimental results show that the steganographic framework guarantees the quality of steganography while its relative steganographic capacity reaches 1.Besides,Peak Signal-to-Noise Ratio(PSNR)and Structural Similarity Index(SSIM)values can reach 42 dB and 0.94,respectively.
基金supported by the National Natural Science Foundation of China (No. 61902236)Fundamental Research Funds for the Central Universities (No. JB210311).
文摘Mobile bike-sharing services have been prevalently used in many cities as an important urban commuting service and a promising way to build smart cities,especially in the new era of 5G and Internet-of-Things(IoT)environments.A mobile bike-sharing service makes commuting convenient for people and imparts new vitality to urban transportation systems.In the real world,the problems of no docks or no bikes at bike-sharing stations often arise because of several inevitable reasons such as the uncertainty of bike usage.In addition to pure manual rebalancing,in several works,attempts were made to predict the demand for bikes.In this paper,we devised a bike-sharing service with highly accurate demand prediction using collaborative computing and information fusion.We combined the information of bike demands at different time periods and the locations between stations and proposed a dynamical clustering algorithm for station clustering.We carefully analyzed and discovered the group of features that impact the demand of bikes,from historical bike-sharing records and 5G IoT environment data.We combined the discovered information and proposed an XGBoost-based regression model to predict the rental and return demand.We performed sufficient experiments on two real-world datasets.The results confirm that compared to some existing methods,our method produces superior prediction results and performance and improves the availability of bike-sharing service in 5G IoT environments.
基金Supported by the NSFC(No.60434020,60572051)Science and Technology Key Item of Ministry of Education of the PRC( No.205-092)the ZJNSF(No. R106745)
文摘This paper presents a new idea, named as modeling multisensor-heterogeneous information, to incorporate the fuzzy logic methodologies with mulitsensor-multitarget system under the framework of random set theory. Firstly, based on strong random set and weak random set, the unified form to describe both data (unambiguous information) and fuzzy evidence (uncertain information) is introduced. Secondly, according to signatures of fuzzy evidence, two Bayesian-markov nonlinear measurement models are proposed to fuse effectively data and fuzzy evidence. Thirdly, by use of "the models-based signature-matching scheme", the operation of the statistics of fuzzy evidence defined as random set can be translated into that of the membership functions of relative point state variables. These works are the basis to construct qualitative measurement models and to fuse data and fuzzy evidence.
基金National Natural Science Foundation of China(12161013)Research Projects of Guizhou University of Commerce in 2024。
文摘The modifiedλ-differential Lie-Yamaguti algebras are considered,in which a modifiedλ-differential Lie-Yamaguti algebra consisting of a Lie-Yamaguti algebra and a modifiedλ-differential operator.First we introduce the representation of modifiedλ-differential Lie-Yamaguti algebras.Furthermore,we establish the cohomology of a modifiedλ-differential Lie-Yamaguti algebra with coefficients in a representation.Finally,we investigate the one-parameter formal deformations and Abelian extensions of modifiedλ-differential Lie-Yamaguti algebras using the second cohomology group.
基金Space Optoelectronic Measurement and Perception Lab(LabSOMP-2023-07)the National Natural Science Foundation ofChina(42241147)+1 种基金the State Key Laboratory of Geo-Information Engineering(SKLGIE2021-Z-3-1)and the Open Program of Collaborativeinnovation Center of Geo-information(2023C002)。
文摘This paper proposed an efficient method of image overlapping relationship analysis based on spatial index of KD tree fast search for disordered and large-scale asteroid images.In this study,the image data from asteroid exploration missions such as Bennu,Vesta,and Ryugu were used for experiments,and the proposed image matching pairs determination algorithm was comprehensively compared with the corresponding modules of USGS ISIS in order to evaluate its performance in terms of efficiency and accuracy.The results show that when processing more than a thousand images,the proposed method greatly improves the efficiency of acquiring image matching pairs while ensuring the correctness of image overlapping relationships and accuracy of bundle adjustment.At the same time,according to the obtained image matching pairs,images that meet the requirements of Stereo Photoclinometry can be quickly selected,effectively improving the quality of 3D reconstruction models of asteroid images.
基金supported in Natural Science Foundation of Shandong Province,China(ZR2013FM018)。
文摘This paper aims to enhance the array Beamforming(BF) robustness by tackling issues related to BF weight state estimation encountered in Constant Modulus Blind Beamforming(CMBB). To achieve this, we introduce a novel approach that incorporates an L1-regularizer term in BF weight state estimation. We start by explaining the CMBB formation mechanism under conditions where there is a mismatch in the far-field signal model. Subsequently, we reformulate the BF weight state estimation challenge using a method known as variable-splitting, turning it into a noise minimization problem. This problem combines both linear and nonlinear quadratic terms with an L1-regularizer that promotes the sparsity. The optimization strategy is based on a variable-splitting method, implemented using the Alternating Direction Method of Multipliers(ADMM). Furthermore, a variable-splitting framework is developed to enhance BF weight state estimation, employing a Kalman Smoother(KS) optimization algorithm. The approach integrates the Rauch-TungStriebel smoother to perform posterior-smoothing state estimation by leveraging prior data. We provide proof of convergence for both linear and nonlinear CMBB state estimation technology using the variable-splitting KS and the iterated extended Kalman smoother. Simulations corroborate our theoretical analysis, showing that the proposed method achieves robust stability and effective convergence, even when faced with signal model mismatches.
基金supported by the National Natural Science Foundation of China(52403403)Guizhou Provincial Basic Research Program(Natural Science)(Qian ke he ji chu-ZK2024 YiBan 095)。
文摘Er^(3+)-doped BaLaGaO_(4)green phosphors was synthesized through a high-temperature solid-state reaction technique.The phase structure and morphology test results of the phosphor indicate that the BaLaGaO_(4)material was successfully synthesized and Er^(3+)ions were successfully doped into the main lattice.This doping does change the basic structure of the crystal.BaLaGaO_(4):Er^(3+)phosphor exhibits bright green emission centered at 545 nm when excited by 381 nm ultraviolet light or 980 nm near-infrared light.The optimal doping concentration is found to be x=0.04.To quantify the temperature sensitivity of the phosphor,the fluorescence intensity ratio method was used.Within the temperature range of 298-473 K,the maximum relative sensitivities are 1.35%/K(298 K,381 nm)and 1.45%/K(298 K,980 nm),respectively.The maximum absolute sensitivities are 0.67%/K(473 K,381 nm)and 0.69%/K(473 K,980 nm),respectively.Finally,white light-emitting diodes(WLEDs)with a high colour index of Ra=82and a relatively low correlated colour temperature of CCT=5064 K are obtained by integrating the synthesized BaLaGaO_(4):0.04Er^(3+)green phosphor into warm WLEDs devices.These results suggest that Er^(3+)-activated BaLaGaO_(4)multifunctional phosphors hold considerable promise in the areas of optical temperature sensing and WLEDs phosphor conversion.
基金supported in part by the National Natural Science Foundation of China under Grant No.62172280in part by the Key Scientific Research Projects of Colleges and Universities in Henan Province,China under Grant No.23A520006in part by Henan Provincial Science and Technology Research Project under Grant No.222102210199.
文摘Deep neural network(DNN)models have achieved remarkable performance across diverse tasks,leading to widespread commercial adoption.However,training high-accuracy models demands extensive data,substantial computational resources,and significant time investment,making them valuable assets vulnerable to unauthorized exploitation.To address this issue,this paper proposes an intellectual property(IP)protection framework for DNN models based on feature layer selection and hyper-chaotic mapping.Firstly,a sensitivity-based importance evaluation algorithm is used to identify the key feature layers for encryption,effectively protecting the core components of the model.Next,the L1 regularization criterion is applied to further select high-weight features that significantly impact the model’s performance,ensuring that the encryption process minimizes performance loss.Finally,a dual-layer encryption mechanism is designed,introducing perturbations into the weight values and utilizing hyperchaotic mapping to disrupt channel information,further enhancing the model’s security.Experimental results demonstrate that encrypting only a small subset of parameters effectively reduces model accuracy to random-guessing levels while ensuring full recoverability.The scheme exhibits strong robustness against model pruning and fine-tuning attacks and maintains consistent performance across multiple datasets,providing an efficient and practical solution for authorization-based DNN IP protection.
文摘Lunar wrinkle ridges are an important stress geological structure on the Moon, which reflect the stress state and geological activity on the Moon. They provide important insights into the evolution of the Moon and are key factors influencing future lunar activity, such as the choice of landing sites. However, automatic extraction of lunar wrinkle ridges is a challenging task due to their complex morphology and ambiguous features. Traditional manual extraction methods are time-consuming and labor-intensive. To achieve automated and detailed detection of lunar wrinkle ridges, we have constructed a lunar wrinkle ridge data set, incorporating previously unused aspect data to provide edge information, and proposed a Dual-Branch Ridge Detection Network(DBR-Net) based on deep learning technology. This method employs a dual-branch architecture and an Attention Complementary Feature Fusion module to address the issue of insufficient lunar wrinkle ridge features. Through comparisons with the results of various deep learning approaches, it is demonstrated that the proposed method exhibits superior detection performance. Furthermore, the trained model was applied to lunar mare regions, generating a distribution map of lunar mare wrinkle ridges;a significant linear relationship between the length and area of the lunar wrinkle ridges was obtained through statistical analysis, and six previously unrecorded potential lunar wrinkle ridges were detected. The proposed method upgrades the automated extraction of lunar wrinkle ridges to a pixel-level precision and verifies the effectiveness of DBR-Net in lunar wrinkle ridge detection.
基金supported by the CCF-NSFOCUS‘Kunpeng’Research Fund(CCF-NSFOCUS2024012).
文摘In recent years,with the rapid development of software systems,the continuous expansion of software scale and the increasing complexity of systems have led to the emergence of a growing number of software metrics.Defect prediction methods based on software metric elements highly rely on software metric data.However,redundant software metric data is not conducive to efficient defect prediction,posing severe challenges to current software defect prediction tasks.To address these issues,this paper focuses on the rational clustering of software metric data.Firstly,multiple software projects are evaluated to determine the preset number of clusters for software metrics,and various clustering methods are employed to cluster the metric elements.Subsequently,a co-occurrence matrix is designed to comprehensively quantify the number of times that metrics appear in the same category.Based on the comprehensive results,the software metric data are divided into two semantic views containing different metrics,thereby analyzing the semantic information behind the software metrics.On this basis,this paper also conducts an in-depth analysis of the impact of different semantic view of metrics on defect prediction results,as well as the performance of various classification models under these semantic views.Experiments show that the joint use of the two semantic views can significantly improve the performance of models in software defect prediction,providing a new understanding and approach at the semantic view level for defect prediction research based on software metrics.
基金supported by the National Natural Science Foundation of China(62122064,62331021,62371410)the Natural Science Foundation of Fujian Province of China(2023J02005 and 2021J011184)+1 种基金the President Fund of Xiamen University(20720220063)the Nanqiang Outstanding Talents Program of Xiamen University.
文摘Magnetic resonance imaging(MRI)plays an important role in medical diagnosis,generating petabytes of image data annually in large hospitals.This voluminous data stream requires a significant amount of network bandwidth and extensive storage infrastructure.Additionally,local data processing demands substantial manpower and hardware investments.Data isolation across different healthcare institutions hinders crossinstitutional collaboration in clinics and research.In this work,we anticipate an innovative MRI system and its four generations that integrate emerging distributed cloud computing,6G bandwidth,edge computing,federated learning,and blockchain technology.This system is called Cloud-MRI,aiming at solving the problems of MRI data storage security,transmission speed,artificial intelligence(AI)algorithm maintenance,hardware upgrading,and collaborative work.The workflow commences with the transformation of k-space raw data into the standardized Imaging Society for Magnetic Resonance in Medicine Raw Data(ISMRMRD)format.Then,the data are uploaded to the cloud or edge nodes for fast image reconstruction,neural network training,and automatic analysis.Then,the outcomes are seamlessly transmitted to clinics or research institutes for diagnosis and other services.The Cloud-MRI system will save the raw imaging data,reduce the risk of data loss,facilitate inter-institutional medical collaboration,and finally improve diagnostic accuracy and work efficiency.
基金supported in part by Key Scientific Research Projects of Colleges and Universities in Anhui Province(2022AH051921)Science Research Project of Bengbu University(2024YYX47pj,2024YYX48pj)+8 种基金Anhui Province Excellent Research and Innovation Team in Intelligent Manufacturing and Information Technology(2023AH052938)Big Data and Machine Learning Research Team(BBXYKYTDxj05)Funding Project for the Cultivation of Outstanding Talents in Colleges and Universities(gxyqZD2021135)the Key Scientific Research Projects of Anhui Provincial Department of Education(2022AH051376)Start Up Funds for Scientific Research of High-Level Talents of Bengbu University(BBXY2020KYQD02)Scientific Research and Development Fund of Suzhou University(2021fzjj29)Research on Grain Logistics Data Processing and Safety Issues(ALAQ202401017)the Open Fund of State Key Laboratory of Tea Plant Biology and Utilization(SKLTOF20220131)funded by the Ongoing Research Funding Program(ORF-2025-102),King Saud University,Riyadh,Saudi Arabia.
文摘With the advancements of the next-generation communication networking and Internet ofThings(IoT)technologies,a variety of computation-intensive applications(e.g.,autonomous driving and face recognition)have emerged.The execution of these IoT applications demands a lot of computing resources.Nevertheless,terminal devices(TDs)usually do not have sufficient computing resources to process these applications.Offloading IoT applications to be processed by mobile edge computing(MEC)servers with more computing resources provides a promising way to address this issue.While a significant number of works have studied task offloading,only a few of them have considered the security issue.This study investigates the problem of spectrum allocation and security-sensitive task offloading in an MEC system.Dynamic voltage scaling(DVS)technology is applied by TDs to reduce energy consumption and computing time.To guarantee data security during task offloading,we use AES cryptographic technique.The studied problem is formulated as an optimization problem and solved by our proposed efficient offloading scheme.The simulation results show that the proposed scheme can reduce system cost while guaranteeing data security.
基金supported by Key Science and Technology Program of Henan Province,China(Grant Nos.242102210147,242102210027)Fujian Province Young and Middle aged Teacher Education Research Project(Science and Technology Category)(No.JZ240101)(Corresponding author:Dong Yuan).
文摘Vehicle Edge Computing(VEC)and Cloud Computing(CC)significantly enhance the processing efficiency of delay-sensitive and computation-intensive applications by offloading compute-intensive tasks from resource-constrained onboard devices to nearby Roadside Unit(RSU),thereby achieving lower delay and energy consumption.However,due to the limited storage capacity and energy budget of RSUs,it is challenging to meet the demands of the highly dynamic Internet of Vehicles(IoV)environment.Therefore,determining reasonable service caching and computation offloading strategies is crucial.To address this,this paper proposes a joint service caching scheme for cloud-edge collaborative IoV computation offloading.By modeling the dynamic optimization problem using Markov Decision Processes(MDP),the scheme jointly optimizes task delay,energy consumption,load balancing,and privacy entropy to achieve better quality of service.Additionally,a dynamic adaptive multi-objective deep reinforcement learning algorithm is proposed.Each Double Deep Q-Network(DDQN)agent obtains rewards for different objectives based on distinct reward functions and dynamically updates the objective weights by learning the value changes between objectives using Radial Basis Function Networks(RBFN),thereby efficiently approximating the Pareto-optimal decisions for multiple objectives.Extensive experiments demonstrate that the proposed algorithm can better coordinate the three-tier computing resources of cloud,edge,and vehicles.Compared to existing algorithms,the proposed method reduces task delay and energy consumption by 10.64%and 5.1%,respectively.
基金supported by the National Natural Science Foundation of China Grant(No.61972133)Project of Leading Talents in Science and Technology Innovation for Thousands of People Plan in Henan Province Grant(No.204200510021)the Key Research and Development Plan Special Project of Henan Province Grant(No.241111211400).
文摘At inference time,deep neural networks are susceptible to backdoor attacks,which can produce attackercontrolled outputs when inputs contain carefully crafted triggers.Existing defense methods often focus on specific attack types or incur high costs,such as data cleaning or model fine-tuning.In contrast,we argue that it is possible to achieve effective and generalizable defense without removing triggers or incurring high model-cleaning costs.Fromthe attacker’s perspective and based on characteristics of vulnerable neuron activation anomalies,we propose an Adaptive Feature Injection(AFI)method for black-box backdoor detection.AFI employs a pre-trained image encoder to extract multi-level deep features and constructs a dynamic weight fusionmechanism for precise identification and interception of poisoned samples.Specifically,we select the control samples with the largest feature differences fromthe clean dataset via feature-space analysis,and generate blended sample pairs with the test sample using dynamic linear interpolation.The detection statistic is computed by measuring the divergence G(x)in model output responses.We systematically evaluate the effectiveness of AFI against representative backdoor attacks,including BadNets,Blend,WaNet,and IAB,on three benchmark datasets:MNIST,CIFAR-10,and ImageNet.Experimental results show that AFI can effectively detect poisoned samples,achieving average detection rates of 95.20%,94.15%,and 86.49%on these datasets,respectively.Compared with existing methods,AFI demonstrates strong cross-domain generalization ability and robustness to unknown attacks.
基金co-supported by the Beijing Natural Science Foundation of China (No. 4194074)the National Key R&D Program of China (No. 2017YFC1600605)+1 种基金the Shandong Provincial Natural Science Foundation of China (No. ZR2018BF016)the Beijing Municipal Education Commission Research Program-General Project of China (No. KM201910011011)
文摘Multirotor has been applied to many military and civilian mission scenarios. From the perspective of reliability, it is difficult to ensure that multirotors do not generate hardware and software failures or performance anomalies during the flight process. These failures and anomalies may result in mission interruptions, crashes, and even threats to the lives and property of human beings.Thus, the study of flight reliability problems of multirotors is conductive to the development of the drone industry and has theoretical significance and engineering value. This paper proposes a reliable flight performance assessment method of multirotors based on an Interacting Multiple Model Particle Filter(IMMPF) algorithm and health degree as the performance indicator. First, the multirotor is modeled by the Stochastic Hybrid System(SHS) model, and the problem of reliable flight performance assessment is formulated. In order to solve the problem, the IMMPF algorithm is presented to estimate the real-time probability distribution of hybrid state of the established SHS-based multirotor model, since it can decrease estimation errors compared with the standard interacting multiple model algorithm based on extended Kalman filter. Then, the reliable flight performance is assessed with health degree based on the estimation result. Finally, a case study of a multirotor suffering from sensor anomalies is presented to validate the effectiveness of the proposed method.
基金co-supported by the China Postdoctoral Science Foundation (No.2013M541035)the National Natural Science Foundation of China (No.61301025)
文摘Radio frequency interference（RFI） is becoming more and more frequently, which makes it an important issue in SAR imaging.RFI presented in synthetic aperture radar either on purpose or inadvertent will distort the useful SAR echoes, thus degrade the SAR image quality.To resolve this issue, a long time study was carried out to study the characteristic of the RFI through the RFIaffected spaceborne and airborne SAR data.Based on the narrow band nature of RFI, this paper proposes a new process which contains both RFI detection and RFI suppression.A useful subband spectral kurtosis detector is first used to detect RFI, and then its results are used for RFI suppression.The proposed process has two advantages： one is the economization on the compute time for unnecessary interference suppression when no RFI existed; the other is improving the performance of the suppression method with knowing the exact position where RFI is.Moreover, the previous RFI suppression method––subband spectral cancelation（SSC） is supplemented and perfected.The subband division step is also elaborated detail in this paper.The experiment results show that the subband spectral kurtosis detector exhibits good performance in recognizing both weak and narrow-band RFI.In addition, the validity of the SSC method with subband spectral kurtosis detector is also validated on the real SAR echoes.
