Local Edge Research Articles

Topological phase transitions and switchable edge states for the non-Hermitian Aubry-André-Harper model in photonic crystals.

The Aubry-André-Harper (AAH) model with imaginary periodic or quasiperiodic modulations could regulate the local properties of the system. In this work, we apply the non-Hermitian AAH potential field to the photonic system induced by gain and loss. It is found that the potential field with different parameters will cause the system to exhibit different local properties and induce different localized edge states under the open boundary conditions. Moreover, we investigated the influence of non-Hermitian AAH potential parameters on the edge states of the Haldane model under periodic boundary conditions. It is shown that two-dimensional magnetic photonic crystals can undergo topological transitions and switchable edge states by tuning the potential field parameters. This offers a versatile method for controlling optical flow in photonic platforms.

ADMNet: adaptive deformable convolution large model combining multi-level progressive fusion for Building Change Detection

ABSTRACT Building Change Detection (BCD) based on high-resolution Remote Sensing Images (RSI) simplifies urban surface monitoring. Nevertheless, the mainstream detection methods utilizing traditional convolution and attention mechanisms are often prone to errors due to the loss of edge detail information and underutilization of global context information. To address these issues, this paper presents a large model, namely ADMNet, which is built on adaptive deformable convolution and is designed to handles various types of building change information. First, we propose a Siamese neural network based on adaptive deformable convolution (ADC) modules. The ADC module incorporates spatial offset parameters into convolutional kernel sampling and mapping weights to capture irregularly varying edge features for local adaptive receptive fields. Second, we utilize a large model semantically driven to enhance model context awareness and construct long-range feature dependencies from multi-scale edge information, which are then integrated with locally adaptive edge structure features to achieve accurate edge localization. Furthermore, we design a Multi-Level Progressive Feature Fusion (MLPFF) module that enhances feature characterization capabilities to ensure internal integrity and improves model detection performance by integrating a priori knowledge from large-model transfer learning. To evaluate the effectiveness and generalizability of ADMNet, we conduct comparative experiments with current mainstream methods on two building datasets, LEVIR-CD and WHU-CD, and a land cover dataset, SYSU-CD. The results show that ADMNet outperforms all comparative methods. The source code is available at https://github.com/spaceYu180/ADMNet.

The novel resolvability parameter, local edge partition dimension of graphs

The novel resolvability parameter, local edge partition dimension of graphs

Heterogeneous Edge Computing for Molecular Property Prediction with Graph Convolutional Networks

Graph-based neural networks have proven to be useful in molecular property prediction, a critical component of computer-aided drug discovery. In this application, in response to the growing demand for improved computational efficiency and localized edge processing, this paper introduces a novel approach that leverages specialized accelerators on a heterogeneous edge computing platform. Our focus is on graph convolutional networks, a leading graph-based neural network variant that integrates graph convolution layers with multi-layer perceptrons. Molecular graphs are typically characterized by a low number of nodes, leading to low-dimensional dense matrix multiplications within multi-layer perceptrons—conditions that are particularly well-suited for Edge TPUs. These TPUs feature a systolic array of multiply–accumulate units optimized for dense matrix operations. Furthermore, the inherent sparsity in molecular graph adjacency matrices offers additional opportunities for computational optimization. To capitalize on this, we developed an FPGA GFADES accelerator, using high-level synthesis, specifically tailored to efficiently manage the sparsity in both the graph structure and node features. Our hardware/software co-designed GCN+MLP architecture delivers performance improvements, achieving up to 58× increased speed compared to conventional software implementations. This architecture is implemented using the Pynq framework and TensorFlow Lite Runtime, running on a multi-core ARM CPU within an AMD/Xilinx Zynq Ultrascale+ device, in combination with the Edge TPU and programmable logic.

Safety After Dark: A Privacy Compliant and Real-Time Edge Computing Intelligent Video Analytics for Safer Public Transportation.

Public transportation systems play a vital role in modern cities, but they face growing security challenges, particularly related to incidents of violence. Detecting and responding to violence in real time is crucial for ensuring passenger safety and the smooth operation of these transport networks. To address this issue, we propose an advanced artificial intelligence (AI) solution for identifying unsafe behaviours in public transport. The proposed approach employs deep learning action recognition models and utilises technologies like NVIDIA DeepStream SDK, Amazon Web Services (AWS) DirectConnect, local edge computing server, ONNXRuntime and MQTT to accelerate the end-to-end pipeline. The solution captures video streams from remote train stations closed circuit television (CCTV) networks, processes the data in the cloud, applies the action recognition model, and transmits the results to a live web application. A temporal pyramid network (TPN) action recognition model was trained on a newly curated video dataset mixing open-source resources and live simulated trials to identify the unsafe behaviours. The base model was able to achieve a validation accuracy of 93% when trained using open-source dataset samples and was improved to 97% when live simulated dataset was included during the training. The developed AI system was deployed at Wollongong Train Station (NSW, Australia) and showcased impressive accuracy in detecting violence incidents during an 8-week test period, achieving a reliable false-positive (FP) rate of 23%. While the AI correctly identified 30 true-positive incidents, there were 6 cases of false negatives (FNs) where violence incidents were missed during the rainy weather suggesting more data in the training dataset related to bad weather. The AI model's continuous retraining capability ensures its adaptability to various real-world scenarios, making it a valuable tool for enhancing safety and the overall passenger experience in public transport settings.

Enhanced Aiot Multi‐Modal Fusion for Human Activity Recognition in Ambient Assisted Living Environment

ABSTRACTMethodologyHuman activity recognition (HAR) has emerged as a fundamental capability in various disciplines, including ambient assisted living, healthcare, human‐computer interaction, etc. This study proposes a novel approach for activity recognition by integrating IoT technologies with Artificial Intelligence and Edge Computing. This work presents a fusion HAR approach that combines data readings from wearable sensors such as accelerometer and gyroscope sensors and Images captured by vision‐based sensors such as cameras incorporating the capabilities of Long Short‐Term Memory (LSTM) and Convolutional Neural Network (CNN) models. The aim of fusing these models is to capture and extract the temporal and spatial information, improving the accuracy and resilience of activity identification systems. The work uses the CNN model to find spatial features from the images that represent the contextual information of the activities and the LSTM model for processing sequential accelerometer and gyroscope sensor data to extract the temporal dynamics from the human activities.ResultsThe performance of our fusion approach is evaluated through different experiments using varying parameters and applies the best‐suited parameters for our model. The results demonstrate that the fusion of LSTM and CNN models outperforms standalone models and traditional fusion methods, achieving an accuracy of 98%, which is almost 9% higher than standalone models.ConclusionThe fusion of LSTM and CNN models enables the integration of complementary information from both data sources, leading to improved performance. The computation tasks are performed at the local edge device resulting to enhanced privacy and reduced latency. Our approach greatly impacts real‐world applications where accurate and reliable HAR systems are essential for enhancing human‐machine interaction and monitoring human activities in various domains.

How to Combine SHJ Cell-Edge Passivation and Module Reliability?

The use of cut-cells in most of the current modules is now the norm to maximise the final product performances. But the integration of such new cell configurations also comes with new challenges, especially if small cell size dimensions such as shingle are considered. Indeed, specific edge passivation processes are often used to recover for the cut-cell losses. Alumina Oxide or Polymer deposition are the most studied approaches, with in general promising benefits proven at cell level. However, only few communications after module integration are available, in particular about the impact of these additional layers on final module reliability. We show for example that it is crucial to avoid direct deposition of AlOX over the cell metal pattern in the future interconnection area, as TC (Thermal Cycling) resilience of such modules is clearly degraded. Proper edge localization of the AlOX layer is needed to recover the initial reliability behaviour. To preserve the interconnection quality, three different approaches are investigated: (1) edge localisation of the layer by wafer stacking during deposition (2) introduction of adapted copper plating solutions allowing metal growth through the passivation layer (3) edge passivation process directly applied on final strings just before lamination.

Nonlinear switch and spatial lattice solitons of photonic s-p orbitals.

We develop fs laser-fabricated asymmetric couplers and zig-zag arrays consisting of single- and two-mode waveguides with bipartite Kerr nonlinearity in borosilicate (BK7) glass substrates. The fundamental mode (s orbital) is near resonance with the neighboring higher-order p orbital, causing efficient light transfer at low power. Due to Kerr nonlinearity, the coupler works as an all-optical switch between s and p orbitals. Single- and double-peak spatial lattice solitons of s-p orbitals form due to the bipartite nature of the on-site nonlinearity. We probe highly localized bulk and edge solitons, peaked at the s and p orbitals, spectrally residing in the photonic bandgap. Our work will be important for exploring inter-orbital couplings and nonlinear interactions in intricate photonic devices.

A hybrid classification approach for automatically recognizing COVID-19 using deep transfer learning using chest radiographs

Coronavirus 2019 causes COVID-19, a worldwide epidemic. It endangers millions globally. Early illness detection improves recovery and control. X-ray image processing is used to categorise and identify COVID-19 in the present study. Preprocessing, feature extraction using local binary pattern (LBP) and edge orient histogram (EOH), and classification utilising K-nearest neighbour (KNN), Navie Bayes, support vector machine (SVM), and transfer learning convolution neural networks (CNNs) are some of the stages that are implemented in the process. Other phases in the process include preprocessing, feature extraction, and preprocessing. LBP+KNN, EOH+KNN, LBP+SVM, EOH +SVM, CNN+LBP, and CNN+EOH are the outputs derived from the combinations of feature extraction operators and classifiers. Other possible outcomes are CNN+EOH and CNN+LBP. A total of 4,000 pictures were used as the basis for conducting an analysis of the performance of six different models. In order to train the models, 10-fold cross-validation was used, and their accuracy was measured accordingly. The evaluation results indicate a high level of accuracy in diagnosis, ranging from 90.2% to 97.56%. The CNN+LBP and CNN+EOH models have demonstrated superior performance compared to other models, achieving average accuracies ranging from 96.66% and 98.54%.

HDF-Net: Capturing Homogeny Difference Features to Localize the Tampered Image.

Modern image editing software enables anyone to alter the content of an image to deceive the public, which can pose a security hazard to personal privacy and public safety. The detection and localization of image tampering is becoming an urgent issue to be addressed. We have revealed that the tampered region exhibits homogenous differences (the changes in metadata organization form and organization structure of the image) from the real region after manipulations such as splicing, copy-move, and removal. Therefore, we propose a novel end-to-end network named HDF-Net to extract these homogeny difference features for precise localization of tampering artifacts. The HDF-Net is composed of RGB and SRM dual-stream networks, including three complementary modules, namely the suspicious tampering-artifact prominent (STP) module, the fine tampering-artifact salient (FTS) module, and the tampering-artifact edge refined (TER) module. We utilize the fully attentional block (FLA) to enhance the characterization ability of homogeny difference features extracted by each module and preserve the specifics of tampering artifacts. These modules are gradually merged according to the strategy of "coarse-fine-finer", which significantly improves the localization accuracy and edge refinement. Extensive experiments demonstrate that HDF-Net performs better than state-of-the-art tampering localization models on five benchmarks, achieving satisfactory generalization and robustness.

Localised edge load dependent aeroelastic stability of porous plates with GPL reinforcement under the influence of supersonic flow

Localised edge load dependent aeroelastic stability of porous plates with GPL reinforcement under the influence of supersonic flow

Failure assessment of steel thin-walled cylindrical tanks subjected to shell settlement

Steel thin-walled tanks have been of particular interest in the tank industry from the past to the present due to their unique features. Settlement is one of the most critical conditions that causes widespread damage in reservoirs, and the reason for that is the large volume of reservoirs, improper design, and poor soil condition. According to API 653, differential settlement is the most important cause of failure among them, and local one of the shell edges is a subset of this type of settlement. On this basis, the main objective of this research endeavor is to investigate the nonlinear response of steel storage tanks with and without a roof under local edge settlement. For this purpose, an experimental and numerical study was conducted to investigate the settlement capacity, nonlinear response, and out-of-plane behavior. In this regard, various regulations, and standards, including API 650, API 653, API 625, EEMUA 159 and PIP STE03020, were examined. The findings revealed that the allowable settlement takes place before the initial buckling and snap-through phenomenon occurs in tanks under local edge settlement.

The impact of EEG electrode density on the mapping of cortical activity networks in infants

ObjectiveElectroencephalography (EEG) is widely used for assessing infant's brain activity, and multi-channel recordings support studies on functional cortical networks. Here, we aimed to assess how the number of recording electrodes affects the quality and level of details accessible in studying infant's cortical networks. MethodsDense array EEG recordings with 124 channels from N=20 infants were used as the reference, and lower electrode numbers were subsampled to simulate recording setups with 63, 31, and 19 electrodes, respectively. Cortical activity networks were computed for each recording setup and different frequencies using amplitude and phase correlation measures. The effects of the recording setup were systematically assessed on global, nodal, and edge levels. ResultsCompared to the reference 124-channel recording setup, lowering electrode density affected network measures in a modality- and frequency-specific manner. The global network features were essentially comparable with 63 or 31 channels. However, the analytic reliability of the local network measures, both at nodal and edge levels, was proportional to the electrode density. The low-frequency amplitude correlations were most robust to the number of recording electrodes, whereas higher frequency phase correlation networks were most sensitive to the density of recording electrodes. ConclusionsOur findings suggest strong and predictable effects of recording setup on the network analyses. Higher electrode number supports studies on networks with phase correlations, higher frequency, and finer spatial details. SignificanceThe relationship between the recording setup and reliability of network analyses is essential for the prospective design of research data collection, as well as for guiding analytic strategies when using already collected EEG data from infants.

A hierarchical adaptive federated reinforcement learning for efficient resource allocation and task scheduling in hierarchical IoT network

The increasing demand for processing numerous data from IoT devices in a hierarchical IoT network drives researchers to propose different resource allocation methods in the edge hosts efficiently. Traditional approaches often compromise on one of these aspects: either prioritizing local decision-making at the edge, which lacks global system insights or centralizing decisions in cloud systems, which raises privacy concerns. Additionally, most solutions do not consider scheduling tasks at the same time to effectively complete the prioritized task accordingly. This study introduces the hierarchical adaptive federated reinforcement learning (HAFedRL) framework for robust resource allocation and task scheduling in hierarchical IoT networks. At the local edge host level, a primal–dual update based deep deterministic policy gradient (DDPG) method is introduced for effective individual task resource allocation and scheduling. Concurrently, the central server utilizes an adaptive multi-objective policy gradient (AMOPG) which integrates a multi-objective policy adaptation (MOPA) with dynamic federated reward aggregation (DFRA) method to allocate resources across connected edge hosts. An adaptive learning rate modulation (ALRM) is proposed for faster convergence and to ensure high performance output from HAFedRL. Our proposed HAFedRL enables the effective integration of reward from edge hosts, ensuring the alignment of local and global optimization goals. The experimental results of HAFedRL showcase its efficacy in improving system-wide utility, average task completion rate, and optimizing resource utilization, establishing it as a robust solution for hierarchical IoT networks.

Enhanced Edge Detection through Binary Particle Swarm Optimization and L0 Guided Filtering

Detecting edges holds significant importance in image processing, serving as a fundamental step in numerous computer vision applications. This paper presents an innovative method for performing edge detection by combining Binary Particle Swarm Optimization (BPSO) with L0 Guided Filtering. The proposed method aims to address the challenge of accurately detecting edges in noisy and complex images by leveraging the benefits of both BPSO and L0 guided filtering. The process begins with the initialization of the BPSO algorithm, where binary particles traverse the solution space to optimize parameters critical for edge detection. These optimized parameters are subsequently employed in the L0 guided filtering framework, a sophisticated edge preserving filter known for its ability to maintain fine details while effectively reducing noise. The synergy of BPSO and L0 guided filtering demonstrates improved adaptability to diverse image characteristics, enhancing the overall robustness of edge detection. The binary nature of BPSO allows for efficient exploration of the solution space, facilitating faster convergence to optimal parameters. Concurrently, the L0 guided filtering ensures edge preserving smoothing, contributing to the suppression of unwanted artifacts. Experimental evaluations on benchmark datasets showcase the effectiveness of the proposed method compared to traditional edge detection techniques. The results indicate superior edge localization and reduced sensitivity to noise, highlighting the potential of the BPSO Based Edge Detection under L0 Guided Filtering in real world applications. The presented approach offers a valuable contribution to the advancement of edge detection methodologies, demonstrating its potential for enhancing the performance of computer vision systems in various domains.

Topological Band Engineering of One-Dimensional π-d Conjugated Metal-Organic Frameworks.

One-dimensional (1D) π-d conjugated metal-organic frameworks (c-MOFs) have garnered widespread research interest in chemical energy storage and conversion. In this work, we introduce a universal principle to engineer the topological bands of 1D c-MOFs. Connected by d orbitals of transition metals, two equivalent hidden molecular π orbitals in 1D c-MOFs can generate a staggered hopping within and between the organic ligands, forming Su-Schrieffer-Heeger-shaped 1D topological bands. Guided by this discovery, we investigate the electronic structures of the typical 1D c-MOF assembled from Ni atoms and 2HQDI (QDI = 2,5-diamino-1,4-benzoquinonediimine) precursors (NiQDI) by first-principles calculations, revealing 1D topological bands around the Fermi level. Due to local bonding variations at the QDI terminations, these two hidden molecular π orbitals become atomically bonded but electronically separated at the edge QDI, creating spatially localized in-gap topological edge states at the end of the NiQDI chain. This definitive signature for 1D topological bands is identified through differential conductance spectra in scanning tunneling microscopy measurements. Our results provide conclusive experimental evidence for topological bands in 1D c-MOFs, paving the way for exploring the topological physics in organic materials through frontier molecular orbitals.

Corner and edge states in topological Sierpinski Carpet systems

Fractal lattices, with their self-similar and intricate structures, offer potential platforms for engineering physical properties on the nanoscale and also for realizing and manipulating high order topological insulator states in novel ways. Here we present a theoretical study on localized corner and edge states, emerging from topological phases in Sierpinski Carpet within a $\pi$-flux regime. A topological phase diagram is presented correlating the quadrupole moment with different hopping parameters. Particular localized states are identified following spatial signatures in distinct fractal generations. The specific geometry and scaling properties of the fractal systems can guide the supported topological states types and their associated functionalities. A conductive device is proposed by coupling identical Sierpinski Carpet units providing transport response through projected edge states which carry on the details of the system's topology. Our findings suggest that fractal lattices may also work as alternative routes to tune energy channels in different devices.

CyberAIBot: Artificial Intelligence in an Intrusion Detection System for CyberSecurity in the IoT

The cyber ecosystem presents two interesting properties. Attackers and defenders are normally the same entity; a detailed knowledge of defensive strategies optimises an attack whereas a good defence, based on a layered structure also includes attacks. In addition, Artificial Intelligence (AI) provides new techniques and tools to both attackers and defenders such as Generative Adversarial Networks (GANs) based on Generators and Discriminators for impersonation or Deep Learning (DL) for exhaustive scanning. To address this dilemma, this article presents CyberAIBot: Artificial Intelligence in an Intrusion Detection System for CyberSecurity in the Internet of Things (IoT) aimed at Operational Technology (OT) and Information Technology (IT) network traffic. CyberAIBot is based on a Deep Learning management structure in a private or local edge cloud computing approach where AI makes decisions as close as possible to the source of data. CyberAIBot gradually detects, learns, and adapts to different cyber attacks. In detail, CyberAIBot uses Deep Learning (DL) technical clusters trained in specific attacks and a Deep Learning (DL) management cluster specialises in taking management decisions rather than technical evaluations. This management cluster supervises conflicting classifications from the deep learning technical clusters. CyberAIbot is trained against several datasets and its performance is evaluated between two classification algorithms, the Long Short-Term Memory (LSTM) networks and Support Vector Machines (SVMs). The SVM DL clusters learn faster (average 15x) although the LSTM DL clusters perform better (average 30%). The LSTM DL management cluster performs better than the SVM DL management cluster although it recognises fewer traffic types. The total number of data points analysed by CyberAIBot is 5.52E+08, equivalent to the distance between the planet Earth to the Moon in meters.

Multidirectional Attention Fusion Network for SAR Change Detection

Synthetic Aperture Radar (SAR) imaging is essential for monitoring geomorphic changes, urban transformations, and natural disasters. However, the inherent complexities of SAR, particularly pronounced speckle noise, often lead to numerous false detections. To address these challenges, we propose the Multidirectional Attention Fusion Network (MDAF-Net), an advanced framework that significantly enhances image quality and detection accuracy. Firstly, we introduce the Multidirectional Filter (MF), which employs side-window filtering techniques and eight directional filters. This approach supports multidirectional image processing, effectively suppressing speckle noise and precisely preserving edge details. By utilizing deep neural network components, such as average pooling, the MF dynamically adapts to different noise patterns and textures, thereby enhancing image clarity and contrast. Building on this innovation, MDAF-Net integrates multidirectional feature learning with a multiscale self-attention mechanism. This design utilizes local edge information for robust noise suppression and combines global and local contextual data, enhancing the model’s contextual understanding and adaptability across various scenarios. Rigorous testing on six SAR datasets demonstrated that MDAF-Net achieves superior detection accuracy compared with other methods. On average, the Kappa coefficient improved by approximately 1.14%, substantially reducing errors and enhancing change detection precision.

Machine Learning-Driven RUL Prediction and Uncertainty Quantification for Ball Screw Drives in a Cloud-Ready Maintenance Framework

In today's rapidly evolving industrial landscape, efficient predictive maintenance solutions are essential for minimizing downtime and enhancing productivity. This research introduces an adaptive cloud-based model pipeline for predicting the Remaining Useful Life (RUL) of machine components, specifically ball screws. The pipeline integrates local preprocessing, edge computing, and cloud-based adaptive model training, ensuring data privacy and reducing data transmission volumes. The system classifies wear states using various machine learning mod-els and predicts RUL through regression analysis, incorporating uncertainty quantification for robust maintenance scheduling. The experimental setup includes accelerated degradation of ball screws, with data collected via a three-dimensional accelerometer. Feature extraction and data augmentation techniques are employed to enhance prediction accuracy. Random Forest and Gradient Boosting models demonstrate superior performance, with Random Forest selected for its robustness and uncertainty quantification capabilities. Empirical results indicate high prediction accuracy, with Random Forest achieving up to 91% accuracy in Phase 2. This cloud-ready predictive maintenance framework leverages scalable cloud infrastructure for efficient data processing and real-time updates, offering a practical solution for industrial applications. The proposed approach significantly advances the adoption of digital business models within the manufacturing industry, providing a reliable and efficient tool for predictive maintenance.