Search

Trust detection and node authentication within fog computing for the Vehicle Ad-hoc Networks (VANETs) are used to determine whether the automobiles and other infrastructure elements including Roadside Units (RSUs) are valid and secure before permitting them to function on the system. It is essential for ensuring the privacy and authenticity of automobile networks, particularly for systems that require safety-critical communication. However, the distributed and decentralized design of fog computing increases difficulties in establishing security measures and coordination between various nodes that are responsible for trust detection and authentication. In fog-enabled VANETs, trust and privacy remain a key challenge. To overcome the existing challenges, a new system for node authentication and trust detection in fog computing for VANETs is developed. Initially, node authentication and trust detection in vehicular networks is conducted using Adaptive Deep Neural Networks (ADNN). Verification of the node’s authenticity and evaluating its trust scores will considerably minimize the chance of cyber attacks and fraudulent behavior across the system, thus enhancing the security of the entire system. Node authentication on the VANET model promotes safe interaction between vehicles. Trust detection in VANET guarantees the integrity of information transferred between vehicles. The parameters of the ADNN are optimally tuned with the help of an Enhanced Garter Snake Optimization Algorithm (EGSOA) to enhance the performance. Some of the models focus only on node authentication and do not consider privacy issues. Thus, it affects the users’ identities and personal information. So, in our model after completing the node authentication and trust detection, privacy preservation of data is performed using Optimal Key-aided Data Sanitization (OPDS). Here, the same EGSOA strategy is employed to get the sanitized key to increase security. The effectiveness of this newly developed fog computing framework for VANETs is evaluated against traditional models, with an expectation of achieving superior accuracy.

Image semantic segmentation represents a significant area of research within the field of computer vision. With the advent of deep learning, image semantic segmentation techniques that integrate deep learning have demonstrated superior accuracy compared to traditional image semantic segmentation methods. Recently, the Mamba architecture has demonstrated superior semantic segmentation performance compared to the Transformer architecture, and has consequently become a research focus in this field. Nevertheless, the specifics of the Mamba architecture have remained underexplored in the extant literature. This review provides a comprehensive overview of the latest research progress in deep learning techniques for semantic segmentation. It offers a systematic review of traditional convolutional neural network (CNN)-based architectures and focuses on a series of emerging architectures, including the Transformer architecture, the Mamba architecture, and cutting-edge approaches such as self-supervised learning strategies. For each category, a detailed account is provided of the principal algorithms and techniques employed, together with a report on the performance achieved using datasets commonly used in the field.

The high proportion of distributed photovoltaic and electric vehicle access has a great impact on the distribution network. For hydropower-rich areas, the “hydropower, photovoltaic, storage and charging” collaborative planning method can effectively utilize and tap the flexible adjustment potential of small hydropower and energy storage to solve this problem. First, with the goal of optimizing economic cost, charging distance and system carbon emissions, a “hydropower, photovoltaic, storage and charging” distribution network collaborative optimization planning model is established. Secondly, the trip chain method is used to predict the charging demand of electric vehicles. The power output curve of a typical day is obtained using the K-means clustering algorithm and the hierarchical analysis method. The non-dominated sorting genetic algorithms II (NSGA-II) with elite strategy is used to solve the multi-objective model to obtain the Pareto solution set. The optimal ideal solution is used for screening to obtain the optimal planning scheme under different scenarios. Then, a comprehensive evaluation is carried out on the optimal planning schemes of different scenarios. Finally, a case analysis is carried out using the IEEE 33-bus system as an example. The simulation optimization results show that the proposed method can take into account the multi-dimensional evaluation indicators of the planning scheme, including economy, sociality, low carbon and reliability.

The Wind turbine converters, as the core components of wind energy conversion systems, are crucial for the performance and reliability of the entire wind power generation system. Due to the uncertainty of wind speed and temperature, the converter load is in a state of fluctuation over the long term. Additionally, there is a correlation between the lifetimes of the submodules within the converter, making it difficult for traditional reliability assessment methods of power electronic equipment to accurately reflect the changes in the reliability of the converter. Therefore, a reliability assessment method for the wind power converter system that comprehensively considers the mission profile, and the correlation of submodule lifetimes is proposed. By establishing a wind turbine model and a thermal network model, combined with the analytical life model, the component lifetime of the module under the mission profile is calculated. Secondly, the Weibull distribution expression of the IGBT module component lifetime is obtained using the Monte Carlo method, and a dynamic Copula reliability calculation model for the series system is established, considering the correlation of submodule lifetimes, to assess the reliability of the entire wind power converter system. The study shows that the mission profile has a significant impact on the lifetime of the IGBT module, especially in high wind speed areas and extreme seasonal conditions, where the component lifetime is consumed more severely. Moreover, the constructed reliability calculation model accurately reflects the impact of the dynamic changes in the correlation of submodule lifetimes on the system’s reliability, verifying that considering the correlation of submodule lifetimes is essential for improving the accuracy of the system reliability assessment.

The integration of traditional Chinese medicine (TCM) diagnostics with modern artificial intelligence (AI) techniques has emerged as a promising approach to enhance the objectivity and accuracy of disease assessment. Tongue diagnosis, a non-invasive and unique TCM practice, plays a critical role in evaluating health status but is often limited by the subjective judgment of practitioners. This study addresses these limitations by developing an intelligent tongue diagnosis system using the Cv-Swin Transformer architecture. The system processes a diverse dataset of 5,365 tongue images, classifying them into ten categories based on TCM diagnostic standards. Key findings indicate that the Cv-Swin Transformer model achieves an average accuracy of 87.37% in tongue image classification, demonstrating superior performance compared to traditional models. The system effectively captures complex tongue features related to various diseases, providing precise health assessments and personalized treatment recommendations. This research represents a significant advancement in integrating AI with TCM, offering a robust tool for objective diagnostics and supporting the modernization of traditional practices.

Industry 4.0, where the convergence of digital technology impacts industrial operations and processes, is characterized by cybersecurity resilience. Therefore, in Industry 4.0, Machine Learning (ML) approaches offer enormous potential for strengthening cyber defenses, guaranteeing resistance against cyber attacks, and improving cyber resilience. One of the most significant trends is the rise of ML in cybersecurity. ML’s ability to analyze vast amounts of data and detect threats that human operators will miss positions it as a crucial tool in the cybersecurity arsenal. This survey offers a comprehensive overview and examines how ML supports facets of cybersecurity, including risk evaluation, incident response sharing threat intelligence, intrusion detection, and safeguarding ML models from attacks. This survey discusses existing techniques’ benefits and drawbacks, identifies emerging trends, and proposes research directions by scrutinizing current frameworks, case studies, and methodologies. Furthermore, we discuss several topics, including predictive risk assessment approaches, collaborative threat intelligence sharing platforms, ML-driven intrusion detection models, automated incident response strategies, and techniques for mitigating manipulations in ML models. Furthermore, the survey identifies the applications of language models in enhancing cybersecurity resilience. This article intends to offer an in-depth look at the advancements by drawing on knowledge from academic disciplines. Moreover, the survey aims to inspire concepts and strategies for bolstering cyber resilience in Industry 4.0 environments.

This paper presents a novel design for high-quantum-efficiency and broadband Ge-on-Si photodetectors. The design entails placing a SiO2-filled nano-slit metasurface on a free-standing resonant-cavity enhanced (RCE) Ge structure to achieve high quantum efficiency in a broad spectral range of 1000–1600 nm. This design addresses the limitations of narrow resonance linewidths for conventional RCE devices with Fabry–Perot cavities. The SiO2-filled metasurface introduces diffraction and additional phase matching of the incident light field to horizontal guided modes, thus enhancing light coupling to the active region at non-resonance wavelengths of the Fabry-Perot mode. The device with this design exhibits an average quantum efficiency of 73.28 % in the 1000–1600 nm spectral range and a minimum quantum efficiency of approximately 60%, which is three times higher than a device without the metasurface structure. The quantum efficiency at around 1300 nm reaches 76.65%, and that at around 1550 nm reaches 98.69%. The estimated responsivity is 0.81 A/W and 1.23 A/W, respectively. For high-speed applications, a device comprising a Si spacer below a thin Ge layer of 600 nm thickness is aimed for operation at a bandwidth of 55 GHz. This device exhibits an average of 64.96% quantum efficiency with a minimum quantum efficiency of around 60%. Its peak quantum efficiency reaches 92.61% at around 1300 nm and 94.65% at around 1550 nm. The corresponding estimated responsivity is 0.98 A/W and 1.18 A/W, respectively. Both designs are innovative solutions for high-speed infrared photodetectors of high quantum efficiency over a broad spectral range.

In recent years, deep learning has been effectively used to reconstruct speckled images created by the scattering media. However, when a network with previous data training is used to test new data, it usually exhibits poor performance due to significant differences in new data that differ in scattered media and data types. Retraining the network and collecting a large amount of new data are usually required to enhance performance, but this comes with a high cost and time expense. This article proposes a transfer learning approach to address this issue. We constructed an experimental setup for capturing infrared speckled images, which includes a laser module with a 4f resonant cavity structure capable of adjusting the aperture size of spatial filters. Speckled images are collected and preprocessed at different aperture sizes, then input into a convolutional neural network for training. Transfer learning is employed to study the reconstruction of speckled images between different aperture sizes. Subsequently, to recreate speckle patterns, a speckle reconstruction network named TCGAN (Transformer_CNN_GAN) is proposed. This network is based on pre-training TCGAN with an enormous quantity of data and fine-tuning its model with new data so that it can be adapted quickly with a small sample size. The experimental findings show that TCGAN may achieve good reconstruction without too much new data by means of transfer learning. Moreover, pretraining the network with more complex datasets helps enhance the network’s transfer learning capability. Therefore, conducting research on optical field imaging and restoration methods based on transfer learning holds significant theoretical and practical significance.

Ensuring the security and safety of a real-time operating system (RTOS) is crucial for protecting against potential cyber-attacks. A robust system can provide reliable and uninterrupted operations in the system as well as protect sensitive data and critical functions. To help programmers develop robust programs and systems, multiple secure coding standards have been developed by different organizations. These coding standards offer guidelines for writing secure code to eliminate vulnerabilities and prevent undefined behaviors from happening. The goal of this research is to understand whether several free common RTOSs follow these secure coding standards. We chose MISRA-C:2012 for our experiment because it is a widely recognized coding standard specifically designed for the safe use of the C programming language in critical systems. To investigate whether existing RTOSs adhere to MISRA-C:2012, we utilized a static code analysis tool, Cppcheck, to analyze 16 open-source RTOSs (written in C) for compliance with 153 rules in MISRA-C:2012. The results indicate that most of the RTOSs conform to mandatory rules of MISRA-C. However, quite a few of the required rules are still violated by many of the RTOSs. Some of the violated rules can potentially result in critical issues in the RTOSs, but most of them can be avoided by simply writing better and safer C code. This paper discusses the major issues in the implementation of RTOSs found in our experiment. The analysis of the results will help programmers and researchers understand the primary concerns in RTOSs’ code and learn how to avoid them by implementing better code.

Autonomous driving is a prominent research area, with the primary challenge being the precise perception and interpretation of the surrounding environment. To address issues such as low resolution, dense occlusion, and small targets in complex road scenes, we propose an improved road target detection algorithm, SNCE-YOLO, based on YOLOv5. Firstly, to avoid the loss of feature information during training for low-resolution and small targets, the Space-to-Depth Convolution(SPD-Conv) is introduced to replace the original convolutional layer for retaining global feature information. Secondly, to address the issue of confusing complex backgrounds with detection targets when features are extracted too early and deeply, the Normalization-based Attention Module (NAM) is incorporated into the backbone network without increasing the computational effort, focusing on the channel and spatial information of interest. Thirdly, to enhance the capture and fusion of multi-level feature information, we optimized the neck network’s feature fusion module by implementing the C2f structure. Finally, to reduce the bias in the regression of the occluded target positions, we introduced the EIOU loss function to better define the relationship between the predicted and real frames. The SNCE-YOLO algorithm achieves a mean Average Precision (mAP@0.5) of 91.9% on the KITTI dataset, which represents a 2.6% improvement over the baseline model. Additionally, with a detection speed of 95.24 frames per second (FPS), it meets the requirements for real-time target detection in complex road scenes. Comparisons are also made with other mainstream target detection algorithms to verify the effectiveness of SNCE-YOLO.

In this paper, we investigate the cooperative economic operation control problem of the integrated energy system (IES) composed of a cluster of energy hubs (EHs) without continuous communication. To solve the problem, a dynamic event-triggered communication strategy and corresponding protocols are introduced. Specifically, a distributed EH output controller is designed based on the consensus algorithm. This enables accurate sharing of electricity and heat load power without requiring knowledge of network parameters. Furthermore, the control method ensures that the pressure is kept within an acceptable range and the frequency is restored to the reference value. By taking energy prices into account, the equal incremental principle is applied to develop a control method of devices to minimize the economic cost. Finally, the case studies are provided to substantiate the efficacy of the proposed control approach.

Accurate identification of crop varieties is vital in seed detection, phenotype collection, and scientific breeding. The purity of maize seeds is a key indicator that affects the quality of the seeds. The vigor, physical indicators, and germination rate of different maize varieties seeds are different, which in turn affects the yield and quality of maize. Therefore, the purity of maize varieties must be identified. Traditional identification methods mainly rely on manual identification, which is highly subjective, time-consuming, labor-intensive, and inefficient. For the identification of the Denghai 605 maize variety, three datasets were constructed: germ surface, non-germ surface, and a mixture of germ and non-germ surfaces. A combination of deep learning and convolutional neural network (CNN) was utilized, and an enhanced VGG16 (E-VGG16) model was proposed by adding BN layers and depth concatenation convolution layers. The classification accuracy rates of the proposed E-VGG16 model on the three datasets were 98.13%, 99.09%, and 98.65%, respectively, which were 2.97%, 7.27%, and 2.58% higher than that of the VGG16 model. The minimum values for Macro avg of precision, recall, and F1 score are 0.978261, 0.984127, and 0.980824, respectively. The minimum values for Weighted avg of precision, recall, and F1 score are 0.98212, 0.981308, and 0.981365, respectively, indicating the excellent performance of E-VGG16. This study provides an objective, accurate, time-saving, efficient, and non-destructive method for the purity identification of Denghai 605 and provides a reference for the construction of a crop seed purity identification model based on CNNs.

Under the condition of a small time scale (e.g. second), distributed photovoltaic (PV) power generation output has the problems of strongly fluctuating and difficult to accurately simulate. It affects the control strategy and operation mode of hybrid energy systems. To address this problem, a data-driven small-scale distributed PV plant power output model on a 1-second time scale is proposed for the generation of second-by-second PV power output scenarios in hybrid energy systems. Firstly, this work analyzes the characteristics of PV power output at the 1-second time scale based on the probability distribution of power output fluctuations. Secondly, an index system that characterizes the PV power output fluctuation characteristics at the 1-second time scale is constructed. Then, using the data-driven method, a BP neural network model is constructed to simulate the PV power output at the 1-second time scale. Finally, a simulation is performed using the measured data from the PV plant. The findings demonstrate that compared to PV power output models in seconds based on Pearson systematic random numbers: (1) The correlation coefficient (r) of the proposed model is more than 0.8, in a higher degree of fit; (2) The root mean square error (RMSE) of the proposed model achieves 0.005, generally representing a 37.12% reduction. Overall, both the time scale and model accuracy of this model have deep potential value in PV power output modeling and system regulation.

An actor-critic based distributed deep reinforcement learning approach is proposed to optimize the reactive power of the distribution network under the access of distributed photovoltaics, wind turbines and other power sources. This approach can optimize and dispatch the resources of the distribution network in real time under the change of power output such as distributed photovoltaics and wind turbines, so as to optimize the reactive power of the distribution network. First, this paper builds an optimization model with the objective function of minimizing the reactive power of the distribution network, and considers the operating constraints. Then, the agents of the proposed approach are trained, and the well-trained agents can schedule and optimize the resources of the distribution network in real time. Finally, based on the actual source-load output data in a certain place, reactive power optimization simulation experiments are carried out on the IEEE 33-bus, IEEE 123-bus simulation systems and the actual power distribution system in a region of China. Simulation results show that the proposed distributed deep reinforcement learning approach (DDRLA) can optimize distribution network reactive power online in real time.

Renewable distributed generations are associated with generation intermittency. Exacerbated by the consumption and demand uncertainty and their resulting mismatch, its energy trading suffers similar uncertainty. The situation is severe in the standalone distributed generations (SDG) for lacking transaction access to the utility grid. This paper proposes the energy transaction time determination and minimization algorithm for consumers in the SDG arena. First, blockchain technology is adopted for transaction enhancement and transaction data acquisition. The acquired blockchain data includes the hourly nodes (number of blockchain members), transaction sizes, and corresponding transaction durations. Next, the blockchain-recorded transaction data are fitted using the linear regression (LR) algorithm to obtain their fitting formula. The fitting formula was subsequently optimized in hourly intervals to obtain the optimal transaction time (energy delivery time) using particle swarm optimization (PSO). Finally, the optimal results are presented in a decision tree (DT) to energy consumers in the blockchain platform. Consequently, their transaction decision-making is guided by the result against the inherent transaction time uncertainty. Consumers can thus correctly adjust their transaction habits to suitably adapt to the transaction duration fluctuations in the energy trading arena. Energy transaction delays and transaction costs are consequently minimized leading to greater penetration of renewables and bridging the generation and consumption gap.

Green energy vehicles often use technologies that reduce lower inherent noise. However, adverse weather condition and low visibility at night can cause a glare effect from the headlights of oncoming cars. This poses a major threat to traffic safety. In order to solve this problem, this study initially adopts single frame difference for video frame selection, which reduces the computational load of image processing pipeline. Then, combined with visible and infrared images, this paper uses non-downsampled contourlet transform to achieve glare elimination. Finally, an improved convolutional network is used to detect pedestrians in anti-glare images, and volumetric Kalman filter algorithm is used to track pedestrians. Through these operations, the research establishes a Single-Frame Difference-Based Image Fusion Glare-Resistant Detection System applicable to green energy vehicles. The experimental analysis shows that the designed system can eliminate glare more than 80%, and the pedestrian detection accuracy reaches 95.44%. The constructed system aids green energy vehicles in accurately perceiving their surroundings during nighttime driving, ensuring safe travel.

Underwater Wireless Optical Communication (UWOC) holds significant implications in the realms of marine scientific observation, marine military security, and marine resource utilization. Overcoming the challenges associated with cross-medium information transmission represents a pivotal research frontier within the domain of underwater Internet of Things (IoT). This paper investigates the across water-air-interface ultraviolet light uplink communication in the presence of waves. Different degrees of diffuse-line-of-sight and different sizes of waves are studied in the communication link to assess their impact on transmission performance. Using non-return-to-zero on-off keying (NRZ-OOK) modulation, a data transmission of 17.4 Mbit/s is achieved with a lateral displacement of 5 cm in still water, and 17.1 Mbit/s data transmission is achieved under aligned 12 mm wave amplitude condition. A method is proposed to reduce the dependence on relay in across water-air-interface communication when the optical transmitter and receiver are not aligned, and the effectiveness of using ultraviolet light communication under large wave conditions is demonstrated. Bit error rate and received optical power indicate that increasing wave size and deviation of the transceivers both result in channel attenuation in the water-air-interface channel, thus degrading link performance.

Efficient and rapid analysis of X-ray images is crucial for detecting contraband, including weapons and explosives, at transportation hubs. While existing models like YOLOv8 offer high performance, they often fall short in real-time scenarios due to slow processing speeds. This study proposes SC-Lite, an improved model based on YOLOv8, designed to improve real-time detection of prohibited items in environments with limited computational power, such as embedded systems. We propose a novel acceleration module, the CSPNet Faster Convolution Network Module (C2F_FM), which reduces computational redundancy and optimizes memory usage by focusing standard convolutions on key channels, thereby enhancing efficiency. Furthermore, the Adaptation-BiFPN module improves multi-scale feature fusion with dynamic, content-aware weighting, increasing both the sensitivity and accuracy of detections. Additionally, the LAMP pruning strategy is utilized to optimize the model for deployment in resource-limited environments, significantly reducing the complexity of its architecture. After extensive experimental validation, SC-Lite outperforms all mainstream models in terms of real-time performance and detection accuracy, and the model outperforms existing methods in all metrics. The model achieved reductions of 72%, 62%, and 70% in parameters, computation, and model size, respectively. On the LSIray dataset, SC-Lite boosted the mean average precision (mAP) by 0.3% to 94.3% and enhanced the frame rate from 87 FPS to 136 FPS, an improvement of 49 FPS. For the OPIXray dataset, SC-Lite increased the mAP by 1.5% to 94.8%, with the frame rate rising from 344 FPS to 400 FPS, an improvement of 56 FPS. This study provides an innovative solution to improve the real-time performance and accuracy of security inspection systems.

Siamese network-based trackers have demonstrated competitive performance in the domain of single object tracking. However, their effectiveness is significantly hindered when the target undergoes challenges such as deformation and illumination changes, due to the fixed template features from the first frame. To address this issue, we propose a novel tracking framework called SiamTAR. This framework adaptively updates the current frame’s template by fusing template features from the first frame with updated template features and tracking box features from the previous frame, thereby effectively improving tracking accuracy. Additionally, to reduce the tracker’s attention to redundant information such as similar shapes, colors, and textures near the target, we designed a feature refinement module. This module integrates three attention mechanisms through two parallel branches to capture critical target information, allowing the tracker to ignore some redundant information. To tackle issues of tracking box drift and inaccurate scale estimation during online tracking, we introduce a relocation mechanism. This mechanism corrects the tracking box position by merging the output tracking box features with the template features. Extensive experiments on multiple datasets validate the superior tracking performance of SiamTAR. Specifically, on the GOT-10K dataset, SiamTAR surpasses the current leading Siamese tracker, SiamPW-RBO, by 1.5% in AO and 7.5% in $SR_{0.75}$ metrics, achieving a tracking speed of 26.23 FPS. Source code is available at https://github.com/rkj12345/SiamTAR.

This paper studies the time-dependent cold chain logistics vehicle routing problem considering both traffic congestion and carbon emissions. A cold chain logistics model with time-dependent green vehicle paths with time windows (TDGVRPTW) was developed to fulfil the demands of green logistics and to take comprehensive account of consideration should be given to factors such as road congestion and carbon emissions. The objective of the model is to minimise total costs, which include carbon emission costs, penalty costs, fuel consumption costs, fixed costs, damage costs and refrigeration costs. Two-phase hybrid search algorithm was developed to solve this model. During the initial stage of the algorithm, a dual-population ant colony optimization (DACO) algorithm sharing the optimal individual is employed to optimize the distribution route of the vehicle. During the second phase, an adaptive golden section search (AGSS) algorithm is used to optimise the departure time of the vehicle from the distribution centre to avoid traffic congestion time periods. To validate the effectiveness of the suggested two-phase hybrid search algorithm, it is applied to the improved Solomon benchmark test set. The experimental findings demonstrate that the two-phase hybrid search algorithm can reasonably plan the driving routes and departure times for each vehicle, effectively avoiding peak traffic congestion periods in the city, and reducing the overall delivery cost.

With the evolution of the sixth generation (6G) mobile communication technology, the terahertz (THz) spectrum has attracted much attention in wireless communication applications due to its high bandwidth and low signal transmission delay. However, the introduction of the THz spectrum leads to higher path transmission losses and complex signal attenuation. This makes user sensing and localization in THz massive multiple-input multiple-output (MIMO) systems more challenging. In this paper, we investigate the user sensing and localization problem in THz massive MIMO systems with the assistance of reconfigurable intelligent surfaces (RIS). Firstly, the received signal is modeled as a tensor, and a parallel factor (PARAFAC) method is proposed. The minimum description length (MDL) is then utilized to detect the number of scattering paths in the channel. On this basis, the alternating least squares (ALS) algorithm is employed to estimate the factor matrices, followed by the utilization of a straightforward correlation-based approach to obtain channel parameter information. Finally, the positions of users and scattering points are estimated based on the geometric relationship between channel parameters and location coordinates. The simulation results have verified the effectiveness of the proposed scheme compared to the existing competitive algorithms, and indicate that the proposed scheme exhibits superior parameter estimation performance and can achieve localization accuracy at the decimeter level.

Humans can carry various diseases, some of which are poorly understood and lack comprehensive solutions. Such a disease can exists in human eye that can affect one or both eyes is diabetic retinopathy (DR) which can impair function, vision, and eventually result in permanent blindness. It is one of those complex complexities. Therefore, early detection of DR can significantly reduce the risk of vision impairment by appropriate treatment and necessary precautions. The primary aim of this study is to leverage cutting-edge models trained on diverse image datasets and propose a CNN model that demonstrates comparable performance. Specifically, we employ transfer learning models such as DenseNet121, Xception, Resnet50, VGG16, VGG19, and InceptionV3, and machine learning models such as SVM, and neural network models like (RNN) for binary and multi-class classification. It has been shown that the proposed approach of multi-label classification with softmax functions and categorical cross-entropy works more effectively, yielding perfect accuracy, precision, and recall values. In particular, Xception achieved an impressive 82% accuracy among all the transfer learning models, setting a new benchmark for the dataset used. However, our proposed CNN model shows superior performance, achieving an accuracy of 95.27% on this dataset, surpassing the state-of-the-art Xception model. Moreover, for single-label (binary classifications), our proposed model achieved perfect accuracy as well. Through exploration of these advances, our objective is to provide a comprehensive overview of the leading methods for the early detection of DR. The aim is to discuss the challenges associated with these methods and highlight potential enhancements. In essence, this paper provides a high-level perspective on the integration of deep learning techniques and machine learning models, coupled with explainable artificial intelligence (XAI) and gradient-weighted class activation mapping (Grad-CAM). We present insights into their respective accuracy and the challenges they face. We anticipate that these insights will prove valuable to researchers and practitioners in the field. Our ambition is that this in-depth study and comparison of models will inform and inspire future research endeavors, ultimately leading to enhanced disease detection in medical imaging and thereby assisting healthcare professionals.

Large Language Models (LLMs) excel in interactive chat scenarios due to their advanced conversational abilities. However, their training process invariably exposes them to a diverse range of harmful or toxic content, posing significant challenges in ensuring that LLM responses align with human ethical values. Consequently, the detection and quantification of adverse content remains a paramount issue in contemporary research. In this paper, we introduce the SAFE dataset, a novel resource designed to advance safety assessment research in LLMs. Our dataset extends beyond the binary categorization of content into “safe” and “unsafe”. Drawing upon human interpretations of safety, we further delineate unsafe content into six granular categories: Sensitivity, Harmfulness, Falsehood, Information Corruption, Unnaturalness, and Deviation from Instructions. This refined classification aims to enhance LLMs’ ability to discern unsafe data more accurately. In total, we have created a dataset comprising 52,340 instruction-response pairs, each annotated with safety meta-tags. Additionally, we have compiled expert comparative assessments for these indicators. We developed a multi-expert rating model trained on the SAFE dataset, designed to evaluate the responses of LLMs across various dimensions. This approach highlights the potential of our dataset in the realm of safety assessment for LLMs. The model’s capability to provide multi-faceted evaluations reflects an advanced understanding of the nuanced requirements in LLM response assessment. We believe this dataset represents a valuable resource for the community, contributing to the safe development and deployment of LLMs. Our findings and resources are poised to fuel future research endeavors in this domain.

Software defect identification (SDI) is a key part of improving the quality of software projects and lowering the risks that along with maintenance. It does identify the software defect causes that have not been reached yet to get sufficient results. On the other hand, many researchers have recently developed several models, including NN, ML, DL, advanced CNN, and LSTM, to enhance the effectiveness of defect prediction. Due to an insufficient dataset size, repeated investigations, and no longer appropriate baseline selection, the research on the CNN model was unable to produce reliable results. In addition, XAI a well-known explainability approach creates deep models in computer vision, as well as successfully handles the software defect prediction that is easy for humans to understand. To address these issues, firstly we have used SMOTE for preprocessing which was collected from the NASA repository; categorical and numerical data. Secondly, we have experimented with software defect prediction using 1D-CNN and 2D-CNN named lightweight CNN (LCNN). Subsequently, evaluation we have employed a 100-repetition holdout validation. For the cross-validation setup, we utilized the 1D-CNN model was $20\times 1$ , and for the 2D-CNN model, it was $4\times 5 \times 1$ . After that, the results of the experiment were compared and assessed in terms of accuracy, MSE, and AUC. The result shows that 2D-CNN shows 1.36% better contrast with 1D-CNN. Thirdly, we have conducted research on the identification of software defect features via LIME and SHAP in XAI stand as state-of-the-art techniques. However, we cannot use 2D-CNN because it involves more complex relationships, making it challenging to create transparent explanations. That is why we have realized that 1D-CNN will superior result to explain the root cause of software feature identifications. Finally, LIME provides accurate visualization of software defect features in contrast with SHAP, as well as it helps the stakeholders of the software industry easily find actual root causes of software defect identification.

Graph Pooling technology plays an important role in graph node classification tasks. Sorting pooling technologies maintain large-value units for pooling graphs of varying sizes. However, by analyzing the statistical characteristic of activated units after pooling, we found that a large number of units dropped by sorting pooling are negative-value units that contain representative information and can contribute considerably to the final decision. To maintain more representative information, we proposed a novel pooling technology, called Geometric Pooling (GP), containing the unique node features with negative values by measuring the similarity of all node features. We reveal the effectiveness of GP from the entropy reduction view. The experiments were conducted on TUdatasets to show the effectiveness of GP. The results showed that the proposed GP outperforms the SOTA graph pooling technologies by $1\%~\sim ~5\%$ with fewer parameters.

In the realm of renewable energy, wind turbines play a pivotal role in efficiently harnessing wind power, especially in offshore environments, where their significance is amplified. These turbines require vigilant monitoring due to the elevated risk of operational faults. Moreover, vast labeled data is scarce in industrial settings due to the cost associated. Hence, there is a need for fault diagnosis methods that can diagnose precisely with minimal data. This research addresses that problem by proposing an architecture built using prototypical network, few-shot learning, and a modified ultra-lightweight SqueezeNet model specifically made for fault diagnosis. Central to our approach are thermal image datasets captured through infrared (IR) cameras, which enable the detection of subtle temperature variations indicative of faults. The proposed architecture excels in data scarcity. It can swiftly generalize from limited samples, thus reducing the dependence on extensive labeled data and reducing training time. Moreover, the modified model stands out for its highly efficient architecture, featuring 16x lower trainable parameters than SqueezeNet. Despite being ultra-lightweight, our model outperforms the original SqueezeNet by achieving 98% accuracy, 10% higher than the original model, and achieves similar or greater accuracy than other models with significantly more trainable parameters. The proposed architecture achieves optimal computational efficiency while maintaining precise diagnostics. The potential of this technology lies in its ability to be used in real-time fault diagnosis applications on lightweight devices.

The agricultural sector is more vulnerable to the adverse effects of climate change and excessive pesticide application, posing a significant risk to global food security. Accurately predicting crop yields is essential for mitigating these risks and providing information on sustainable agricultural practices. This research presents a novel crop yield prediction system that utilizes a year’s worth of meteorological data, pesticide records, crop yield data, and machine learning techniques. We employed rigorous methods to gather, clean, and enhance data and then trained and evaluated three machine learning models: Gradient Boosting, K-Nearest Neighbors, and Multivariate Logistic Regression. We utilized the GridSearchCV method for hyper-parameter tweaking to identify the most suitable hyper-parameter throughout K-Fold cross-validation, aiming to improve the model’s performance by avoiding overfitting. The remarkable performance of the Gradient Boosting model, with an almost flawless coefficient of determination ( $R^{2}$ ) of 99.99%, demonstrates its promise for precise yield prediction. This research also examined the correlation between projected and actual crop yields and identified the ideal meteorological conditions. It paves the way for data-driven methods in sustainable agriculture and resource distribution, ultimately leading to a more secure future with respect to food availability and resilience to climate change.

Digitization has permeated every aspect of industry, and the integration of immersive digital technology into traditional tourism dance performances has become a core concern in both theoretical and practical fields. However, existing literature lacks sufficient exploration on how the integration of immersive digital technology and tourism dance performances can improve tourist loyalty. Therefore, our research established a theoretical model based on the Stimulus-Organism-Response (SOR) framework and tested our proposed hypothesis through the structural equation model. The research results indicate that authenticity, esthetics, and entertainment have a significant promoting effect on immersive experience and meaningful experience, which have a positive impact on tourists’ revisit intention. In addition, authenticity, esthetics and entertainment promote the improvement of revisiting intention through the mediating effect of flow experience and meaningful experience. This research expands the research context of authenticity theory and experience economy framework, and provides a new research perspective for tourism destination loyalty research. Our conclusion provides theoretical guidance for the sustainable development of tourist destinations.

Camouflaged objects are typically assimilated into their surroundings. Consequently, in contrast to generic object detection/segmentation, camouflaged object detection proves to be considerably more intricate due to the indistinct boundaries and heightened intrinsic similarities between foreground targets and the surrounding environment. Despite the proposition of numerous algorithms that have demonstrated commendable performance across various scenarios, these approaches may still grapple with blurred boundaries, leading to the inadvertent omission of camouflaged targets in challenging scenes. In this paper, we introduce a multi-stage framework tailored for segmenting camouflaged objects through a process of coarse-to-fine refinement. Specifically, our network encompasses three distinct decoders, each fulfilling a unique role in the model. In the initial decoder, we introduce the Bi-directional Locating Module to excavate foreground and background cues, enhancing target localization. The second decoder focuses on leveraging boundary information to augment overall performance, incorporating the Multi-level Feature Fusion Module to generate prediction maps with finer boundaries. Subsequently, the third decoder introduces the Mask-guided Fusion Module, designed to process high-resolution features under the guidance of the second decoder’s results. This approach enables the preservation of structural details and the generation of fine-grained prediction maps. Through the integration of the three decoders, our model effectively identifies and segments camouflaged targets. Extensive experiments are conducted on three commonly used benchmark datasets. The results of these experiments demonstrate that, even without the application of pre-processing or post-processing techniques, our model outperforms 14 state-of-the-art algorithms.

Emotion recognition based on electroencephalography (EEG) signals has garnered substantial attention in recent years and finds extensive applications in the domains of medicine and psychology. However, individual differences in EEG signals pose a challenge to accurate emotion recognition and limit the widespread adoption of such techniques. To address this issue, this study proposes a model that combines random forest weights (RFWs) and four-dimensional convolutional recurrent neural network (4DCRNN) to minimize individual differences and captures emotion-relevant information. By integrating, the proposed model aims to improve the accuracy and generalization capability of emotion recognition. To evaluate the performance of the proposed model, experiments were conducted using the DEAP and SEED datasets. The results demonstrate the effectiveness of the RFW-4DCRNN in emotion recognition. Specifically, the proposed model achieves mean accuracy of 94.98% and 94.21% for Subject-dependent recognition using the DEAP and SEED datasets, respectively. For Subject-independent emotion recognition, the model achieved mean accuracy of 81.70% and 91.12% using two datasets, respectively. The result highlights the capability of the RFW-4DCRNN to effectively recognize emotions and improves generalization performance. Overall, this study presents an approach to addressing individual differences in EEG-based emotion recognition. The RFW-4DCRNN demonstrates promising results in terms of accuracy and generalization capability, offering potential for the advancement and application of emotion recognition techniques.

A novel acquisition scheme of pseudo-noise (PN) codes is proposed for spreading satellite communication systems relying on the proposed multi-voter model of this paper. Based on the proposed model, the acquisition of PN codes can be attributed to a voting and selecting mechanism to pick out the erroneous chips. Although message passing algorithm (MPA) is a feasible algorithm for decoding PN codes, MPA does not get good detection performance due to the limited number of parity check equations in the acquisition scheme. To overcome the negative impact of the limited number of parity check equations, this paper proposes single-voter and multi-voter models, which combine chip and sequence estimation on the basis of the chip-flipping (CF). It is known that the bit-flipping (BF), weighted-bit-flipping (WBF) and other algorithms based on BF are credible for low density parity check (LDPC) codes. Because of the lack of research, these algorithms have not been extended to the detection of PN codes. As the same as BF, the inputs of the proposed CF algorithm are the hard-decision samples. For the proposed CF, there exists an optimal flipping-threshold which is similar to the random weight of the WBF. Owing to the low computation complexity of CF, the unlimited number of parity check equations can be enabled in the voting model. The experimental results show that the detection performance of the proposed method of $N > $ 15 is improved by 2 dB compared with MAP-based method and 4 dB compared with LEAP-based method at the detection probability 99%.

Regional surface deformation resulting from the development and utilization of underground space resources poses a significant threat to the safety of urban areas, and the combination of Synthetic Aperture Radar and Gravity Recovery and Climate Experiment (GRACE) satellite data has become a new means to study the impact of underground space evolution on surface deformation. We combine the Interferometric Synthetic Aperture Radar (InSAR) technology and gravity satellite data to extract information on regional surface deformation and groundwater storage changes in Shanxi Province, to explore the patterns of their temporal and spatial variations, to discover their links with seasonal climate change, to re-conceptualize the laws of the regional water cycle, and to quantify the contribution of multiple fields to the evolution of the surface. Furthermore, we propose a novel multi-source neural network prediction model (LSTM/BP) based on signal decomposition (VMD) and algorithm optimization to handle the complex time series characteristics of groundwater storage. Our findings reveal that groundwater storage in Shanxi Province has been consistently declining, with a monthly deficit rate of approximately 1.05 mm. Additionally, there is a notable spatial variation in the annual rate of change, ranging from −21 to 4 mm/year from north to south. Furthermore, we observe a close relationship between inter-annual and seasonal groundwater storage changes and local rainfall patterns, and we find that regional surface deformation is influenced by these groundwater storage changes. The new prediction model outperforms other models, with a root mean square error of 1.56 mm and a correlation coefficient of more than 0.98 on the test set. The model improves the prediction accuracy of the groundwater reserves in the basin, and it can be used to provide a reference for the comprehensive management of the groundwater in Shanxi Province, the rational development of mineral resources, and other major national needs.

Hypergraph neural networks have gained substantial popularity in capturing complex correlations between data items in multimodal datasets. In this study, we propose a novel approach called the self-supervised hypergraph learning (SHL) framework that focuses on extracting hypergraph features to improve multimodal representation. Our method utilizes a dual embedding strategy and leverages SHL to improve the accuracy and robustness of the model. To achieve this, we employ a hypergraph learning framework to extract global context effectively by capturing rich inter-modal dependencies. Additionally, we introduce a novel self-supervised learning (SSL) component that utilizes the interaction graph data, thereby strengthening the robustness of the model. By jointly optimizing hypergraph feature extraction and SSL, SHL significantly improves the performance of multimodal representation tasks. To validate the effectiveness of our approach, we construct two comprehensive multimodal micro-video recommendation datasets using publicly available data (TikTok and MovieLens-10M). Prior to dataset creation, we meticulously handle invalid entries and outliers and complete missing mode information using external auxiliary sources, such as YouTube. These datasets are made publicly available to the research community for evaluation purposes. Experimental results on the above recommendation datasets demonstrate that the proposed SHL approach outperforms state-of-the-art baselines, highlighting its superior performance in multimodal representation tasks.

The concept of interval-valued intuitionistic fuzzy sets (IVIFSs) presents a compelling and practical framework for modeling real-world problems. In various fields, such as pattern recognition and decision-making, the development of similarity measures tailored to this class holds significant importance. These measures play a pivotal role in the decision-making process involving IVIFSs, as they quantify the extent of similarity between two such sets. In this article, the shortcomings of the existing similarity measures within the framework of IVIFSs are highlighted, and an improved similarity measure is presented. A comparative study validates that this new similarity measure is better than the existing measures in the IVIF environment. This study systematically establishes several essential properties of the novel similarity measure and substantiates its effectiveness through numerical illustrations. Moreover, a comparative assessment is undertaken to validate the efficacy of the recently introduced measure in relation to established metrics, within the context of IVIFSs. To address the evaluation of software quality, a dedicated mechanism is devised, harnessing the proposed IVIFS similarity measure. Furthermore, an innovative production strategy is formulated utilizing the newly defined methodology to determine the optimal approach for the production of a specific product.

This paper deals with the stability of nonlinear multi-agent systems (MASs) with multiple time delays. A class of multi-agent system (MAS) is constructed considering the effects of multiple time delays, nonlinearity, and uncertainty. An integral sliding mode surface is built based on the MAS, and the corresponding sliding mode dynamics are established. Using the Lyapunov method, sufficient conditions are obtained to ensure the sliding mode dynamic asymptotic stability with $H_{\infty }$ performance. Consequently, the sliding mode controller parameter matrix is obtained. Furthermore, a sliding mode control law ensures that the considered system can be driven into the sliding mode surface. Finally, a numerical example is given to demonstrate the usefulness of the theoretical result.

Tumor samples clustering based on subspace segmentation is an effective method to discover cancer subtypes. Accurate and reliable identifications of cancer subtypes are crucial for understanding cancer pathogenesis as well as clinical diagnosis and treatment. Joint learning-based subspace clustering methods utilize the high correlation between data affinity and data segmentation for better clustering results. However, existing joint learning-based methods only provide an approximation of the segmentation cost in the joint optimization term, which could lead to sub-optimal results. To address this problem, we propose an algorithm named joint learning of spectral clustering and low-rank representation based on precise segmentation cost (JLSLPS) for cancer subtype identification. In our method, we impose non-negative and symmetric constraints on the low-rank representation matrix so that the representation coefficient can be equivalent to data affinity for the precise representation of segmentation cost. Therefore, the spectral clustering objective can be represented precisely to guide the learning of data affinity and segmentation more effectively. Finally, we solve the optimization problem of JLSLPS by using the linearized alternating direction method with adaptive penalty. We run JLSLPS on 8 cancer gene expression datasets and used 9 state-of-the-art clustering methods for comparison. Experimental results show that our method can increase ACC by 1.00-7.95% and NMI by 2.7-12.22% compared with the other methods, which proves the superiority of the proposed method.

The core technique of unmanned vehicle systems is the autonomous maneuvering decision, which not only determines the applications of unmanned vehicles but also is the critical technique many countries are competing to develop. Reinforcement Learning (RL) is the potential design method for autonomous maneuvering decision-making systems. Nevertheless, in the face of complex decision-making tasks, it is still challenging to master the optimal policy due to the low learning efficiency caused by the complex environment, high dimensional state, and sparse reward. Inspired by the human learning process from simple to complex, we propose a novel progressive deep RL algorithm for policy optimization in unmanned autonomous decision-making systems in this paper. The proposed algorithm divides the training of the autonomous maneuvering decision into a sequence of curricula with learning tasks from simple to complex. Finally, through the self-play stage, the iterative optimization of the policy is realized. Furthermore, the confrontation environment with two unmanned vehicles with obstacles is analyzed and modeled. Finally, the simulation leads to the one-to-one adversarial tasks demonstrate the effectiveness and applicability of the proposed design algorithm.

The brain network is an effective tool and has been widely used in the field of brain neurodegenerative disease analysis. Due to the high cost of accessing medical image data, efforts have been devoted to investigating data augmentation. However, the brain network containing topological characteristics of non-European space is different from the traditional image data, which makes it challenging to synthesize brain structural connectivity and limits the application of brain network analysis. In this paper, a Hemisphere-separated Cross-connectome Aggregating Learning (HCAL) model is proposed to synthesize realistic and diverse brain structural connectivities. Specifically, the latent representation is transformed from structural connectivity by the graph variational autoencoder (GVAE). To generate more diverse and high-quality structural connectivities, the hemisphere-separated generator with a cross-connectome aggregating mechanism is developed to first learn local topological patterns by splitting the whole brain into inter- and intra-hemispheres, then capture global topological characteristics among all the neighbors for each brain region. Also, the connectivity-aware discriminator is devised to make the adversarial training stable and enhance the disease diagnosis. Evaluations of the public Alzheimer’s Disease Neuroimaging Initiative (ADNI) dataset show that the proposed model generates brain structural networks with higher quality and more diversity than related methods. In addition, the classification performance using the augmented data achieved by our approach achieves an accuracy improvement of over 3% compared with the competing method. The proposed model provides an alternative way to synthesize brain networks for connectivity-based disease analysis.

Cross-modal hashing retrieval has attracted extensive attention due to its low storage requirements as well as high retrieval efficiency. In particular, how to more fully exploit the correlation of different modality data and generate a more distinguished representation is the key to improving the performance of this method. Moreover, Transformer-based models have been widely used in various fields, including natural language processing, due to their powerful contextual information processing capabilities. Based on these motivations, we propose a Transformer-based Distinguishing Strong Representation Deep Hashing (TDSRDH). For text modality, since the sequential relations between words imply semantic relations that are not independent relations, we thoughtfully encode them using a transformer-based encoder to obtain a strong representation. In addition, we propose a triple-supervised loss based on the commonly used pairwise loss and quantization loss. The latter two ensure the learned features and hash-codes can preserve the similarity of the original data during the learning process. The former ensures that the distance between similar instances is closer and the distance between dissimilar instances is farther. So that TDSRDH can generate more discriminative representations while preserving the similarity between modalities. Finally, experiments on the three datasets MIRFLICKR-25K, IAPR TC-12, and NUS-WIDE demonstrated the superiority of TDSRDH over the other baselines. Moreover, the effectiveness of the proposed idea was demonstrated by ablation experiments.

Images captured under foggy or hazy weather conditions are affected by the scattering of atmospheric particles, resulting in decreased contrast and color variation, thereby limiting their practical applications. In recent years, deep learning methods showcase significant advancements in image dehazing. However, the complexity and degradation factors in hazy images challenge the generalization capacity of dehazing methods. This paper comprehensively reviews the recent developments in single-image dehazing techniques based on deep learning. From the perspectives of Convolutional Neural Networks (CNN) and Generative Adversarial Networks (GAN), different models are introduced and classified into four categories: Encoder-Decoder, Multi-Module, Multi-Branch, and Dual-Generative Adversarial Networks. The robustness and effectiveness of deep learning models are analyzed by comparing their performance and model complexity on public datasets. Additionally, limitations of current benchmark datasets and evaluation metrics are identified, and unresolved issues and future research directions are discussed. Our efforts in this paper will serve as a comprehensive reference for future research and call for further development in deep learning-based image dehazing.

The rising deployment of software in automation and the cognitive skills of machines indicate a machine revolution in modern human civilization. Thus, diagnosing and predicting software faults is crucial to software reliability. In this paper, we first preprocessed four real datasets offered by National Aeronautics and Space Administration with twenty-one features using the Synthetic Minority Oversampling Technique and Label Encoding techniques. Subsequently, we experimented with thirteen software fault diagnosis Machine Learning (ML) models, i.e., Random Forest Regression, Linear Regression, Naïve Bayes, Decision Tree Classifier, Logistic Regression, KNeighbors Classifier, AdaBoost, Gradient Boosting Classifier, Gradient Boosting Regression, XGBR Regressor, XGBoost Classifier, Extra Trees Classifier and Support Vectors Machine after that, we compared each ML Model to select the best diagnostic model. Among them, XGBR outperformed, considering the accuracy, mean square error, and R2 score. We also used Explainable Artificial Intelligence (XAI), Local Interpretable Model (LIME), and SHapley Additive exPlanations (SHAP) to determine software fault features. We observed that Number of static invocations (nosi), Depth Inheritance Tree (dit), and Coupling Between Objects (cbo) features are the most affected software faults feature from datasets. For LIME, the average True positive of nosi is 40%, dit is 15%, and cbo is 20%; on the other hand, the SHAP average true positive value of nosi is 36%, cbo is 15%, and the norm true negative value of dit is 5%. Thus, LIME can afford the greatest impact on the model outcomes to identify features that are the most significant reasons for software defects.

Face features, as the most widely adopted and essential biometric characteristic in identity verification and recognition, play a crucial role in ensuring security. However, the significance is also accompanied by various face attacks, posing a great threat to the security of facial recognition systems. Therefore, exploring face anti-spoofing detection holds substantial practical significance. There is existing research on Face Anti-Spoofing (FAS) detection, but there is still potential for enhancing detection performance. In this paper, we propose a face anti-spoofing method, named AR-MLP (Attention ResNet-Multilayer Perceptron), which incorporates self-attention and MLP auxiliary convolution. By replacing the last two “Basic block” layers of the Conv-MLP network with a model combining self-attention and convolution, and adjusting the network architecture and parameters, AR-MLP can capture global features more effectively by suitably integrating the self-attention and convolution layers. We conduct comprehensive evaluations using three authoritative multimodal face anti-spoofing datasets (WMCA, HQ-WMCA, and CASIA-SURF CeFA) for intra-dataset, cross-attack, and cross-race testing. The experimental results demonstrate that AR-MLP outperforms several excellent methods in terms of classification performance and computational overhead.

Automatic visual encoding is frequently employed in automatic visualization tools to automatically map data to visual elements. This paper proposed an automatic visual encoding approach based on deep learning. This approach constructs visual encoding datasets in a more comprehensive and reliable manner to extract and label widely available visualization graphics on the Internet in accordance with three essentials of visualization. The deep learning model is then trained to create a visual encoding model with powerful generalization performance, enabling automated effective visual encoding recommendations for visual designers. The results demonstrated that our approach extends the automatic visual encoding techniques used by existing visualization tools, enhances the functionality and performance of visualization tools, uncovers previously undiscovered data and increases the coverage of data variables.

Parallel coordinates plots are popular tools for high-dimensional data visualization. To alleviate the difficulties caused by the inherent defects that arise when the dimensions are increased, this study attached two-dimensional kernel density scatter plots to parallel coordinates plot, which integrates the Cartesian and parallel coordinate systems, to combine their benefits and conveniently explore the relationships between paired attributes. The collaborative design combining the two visual types was realized through the corelative axis swap interaction method, which allows users to freely exchange parallel coordinate axes while also updating the associated kernel density scatter plots based on the Cartesian coordinate system. In this study, we obtained six cases of real datasets and performed a field trial study to assess the usefulness of our method in assisting users in evaluating correlations between paired attributes. The results show that our method is more useful for analyzing the relationships between paired attributes.

To enable vehicle-to-everything (V2X) communications, both the dedicated short-range communications (DSRC) and the cellular V2X (C-V2X) involved in the radio access technologies (RATs) are experiencing extensive development to support advanced vehicular applications and scenarios. Compared with the C-V2X whose effective reliability and scalability are still to be completely verified, the DSRC maintains its superiority due to the extensive safety-related field trials performed worldwide. Furthermore, IEEE 802.11bd, the next-generation V2X (NGV) standard for the vehicular ad hoc networks (VANETs) in the DSRC, is expected to greatly improve the performance compared to IEEE 802.11p with new physical layer (PHY) and medium access control (MAC) technologies. Hence, aimed to obtain the complete PHY evaluation on IEEE 802.11bd in terms of the packet error rate (PER), the packet reception ratio (PRR), the effective data rate, and the packet inter-arrival time (IAT), the various antenna configurations, vehicle-to-vehicle (V2V) scenarios, packet sizes, and modulation and coding schemes (MCSs) are investigated and compared in a full PHY simulation. The results indicate that the multiple-input multiple-output (MIMO) configuration is the most advantageous technique in decreasing the PER, increasing the PRR and the transmission coverage, elevating the output effective data rate, and reducing the output packet IAT in contrast with the single-input multiple-output (SIMO), multiple-input single-output (MISO), and single-input single-output (SISO) configurations at the farther distance. The urban non-line-of-sight (NLOS) scenario experiences the slightly better quality of communication and is more robust than the highway NLOS scenario. Both the packet size and the MCS need to be selected properly to satisfy the high-reliability, high-throughput, or low-latency requirements.

The information of cropland is obtained efficiently and accurately as the basis for achieving precision agriculture (PA). As the boundary between cropland and other types of land in remote sensing images with different resolutions is fuzzy, the characteristics of cropland are easily confused when extracting cropland, resulting in inaccurate identification and extraction of cropland under large and complex backgrounds and rough localization of marginal areas. We proposed a two-path multiscale attention self-supervised network with the perception of four directions in pixel space, called the multiscale bilateral spatial direction-aware network (MBSDANet), to solve these problems and improve the model’s ability to extract cropland in small samples. One path extracts attentional feature maps by spatial directions to preserve detailed direction-aware information and generate high-resolution features; the other path obtains local-to-global information through pyramid pooling and attention awareness to capture dense multiscale cropland features to separate targets in complex contexts. The features of the two branches at different levels are fused by weighting the multi-aware information. Triangular self-supervised and boundary-aware losses are used to achieve fine segmentation and extraction of cropland in small samples. We tested the extraction method on cropland in Denmark and the Hebei Province of China, demonstrating its effectiveness and generalization. Compared to other neural network models, MBSDANet achieves better accuracy with a precision of 0.9481, an IoU of 0.8937, and an F1 score of 0.9438.

As one fundamental data-mining problem, ANN (approximate nearest neighbor search) is widely used in many industries including computer vision, information retrieval, and recommendation system. $LSH$ (Local sensitive hashing) is one of the most popular hash-based approaches to solve ANN problems. However, the efficiency of operating $LSH$ needs to be improved, as the operations of $LSH$ often involve resource-consuming matrix operations and high-dimensional large-scale datasets. Meanwhile, for resource-constrained devices, this problem becomes more serious. One way to handle this problem is to outsource the heavy computing of high-dimensional large-scale data to cloud servers. However, when a cloud server responsible for computing tasks is untrustworthy, some security issues may arise. In this study, we proposed a cloud server-aided $LSH$ scheme and the application model. This scheme can perform the $LSH$ efficiently with the help of a cloud server and guarantee the privacy of the client’s information. And, in order to identify the improper behavior of the cloud server, we also provide a verification method to check the results returned from the cloud server. Meanwhile, for the implementation of this scheme on resource-constrained devices, we proposed a model for the real application of this scheme. To verify the efficiency and correctness of the proposed scheme, theoretical analysis and experiments are conducted. The results of experiments and theoretical analysis indicate that the proposed scheme is correct, verifiable, secure and efficient.

Retinal image quality assessment (RIQA) is a precondition for diabetic retinopathy screening and diagnosis. Current RIQA network tend to place too much emphasis on performance improvement, ignoring the number of parameters and computational complexity of the network, and the existing publicly available datasets with unclear classification criteria and unbalanced number of different kinds of images, which affect the training of deep neural networks and large area screening of diabetic retinopathy. To solve these problems. A new quality criteria for retinal images is developed, and the Kaggle retinal image quality dataset is re-labeled and reclassified according to the criteria. And a high-performance lightweight RIQA network ADD-Net is proposed based on the DenseNet, which increases the perceptual field and improves pixel utilization by stacking dilated convolutions in DenseLayer, and incorporates attention mechanisms to further optimize the network performance. The proposed ADD-Net has a parametric number of 8.92M, a computational complexity of 2.76GMac, an accuracy of 97.03%, a precision of 97.06%, and a sensitivity of 98.31%, which are better than those of current mainstream methods. A quality criteria for retinal images that combines image quality features, retinal structural features, and DR lesion features was developed. The RIQA network ADD-Net with low number of parameters, less computational complexity and excellent performance is designed.

Data generated from sources such as wearable sensors, medical imaging, personal health records, and public health organizations have resulted in a massive information increase in the medical sciences over the last decade. Advances in computational hardware, such as cloud computing, graphical processing units (GPUs), Field-programmable gate arrays (FPGAs) and tensor processing units (TPUs), provide the means to utilize these data. Consequently, an array of sophisticated Artificial Intelligence (AI) techniques have been devised to extract valuable insights from the extensive datasets in the healthcare industry. Here, we present an overview of recent progress in AI and biosensors in medical and life sciences. We discuss the role of machine learning in medical imaging, precision medicine, and biosensors for the Internet of Things (IoT). We review the latest advancements in wearable biosensing technologies. These innovative solutions employ AI to assist in monitoring of bodily electro-physiological and electro-chemical signals, as well as in disease diagnosis. These advancements exemplify the trend towards personalized medicine, delivering highly effective, cost-efficient, and precise point-of-care treatment.Furthermore, an overview of the advances in computing technologies, such as accelerated AI, edge computing, and federated learning for medical data, are also documented. Finally, we investigate challenges in data-driven AI approaches, the potential issues generated by biosensors and IoT-based healthcare, and the distribution shifts that occur among different data modalities, concluding with an overview of future prospects.

Fingerprint recognition is widely used in daily life. However, the physiological phenomena of human can still fail recognition under dry and low temperature conditions. We present two main approaches for enhancing the accuracy of fingerprint recognition systems in cases of dry fingerprints. The first approach is an image enhancement algorithm with multiple training stages (MTS) for fingerprints at low temperatures, which aims to restore low quality fingerprints to normal quality. The second method is a Cycled Contrastive Ridge Generative Adversarial Network (CCRGAN) learning framework for synthesizing enrolled templates, aimed at improving the accuracy of low-temperature fingerprint matching in authentication systems. For FVC2002 dataset, we have achieved 0.05% for EER on DB1A, which is better than 0.09% for Li et al. (2022) and 0.18% for Wong and Lai (2020) on the MTS algorithm. After adding CCRGAN pseudo low fingerprint, the EER is improved to 0.0009%. As for inference time, our inference time is about 83 times faster than Li et al. (2022) on the FVC2002 DB3B.