Franklin Open最新文献

{"title":"A Comprehensive Deep Learning Approach for Dermoscopic Image Enhancement","authors":"Abdullah Al Mazed ,&nbsp;Md. Faiyaj Ahmed Limon ,&nbsp;Shahidul Haque Thouhid ,&nbsp;Md Fazle Hasan Shiblee ,&nbsp;Shubradeb Das ,&nbsp;Md. Shahid Iqbal ,&nbsp;Debojyoti Biswas","doi":"10.1016/j.fraope.2025.100405","DOIUrl":"10.1016/j.fraope.2025.100405","url":null,"abstract":"<div><div>Image enhancement plays a pivotal role in improving image quality within the field of image processing. In the context of dermoscopic imaging, it serves as a critical and challenging pre-processing step, essential for facilitating accurate automated diagnosis. However, current techniques often struggle to address the diverse range of degradations encountered in real-world scenarios. The primary objective of this study is to propose a robust deep learning approach capable of restoring high-quality images from a wide range of realistic degradation scenarios. To achieve this, we introduce two key contributions: first, EnhanceNet-U, a U-Net architecture modified with a simplified bottleneck and an enhanced decoder path for improved feature restoration; and second, a comprehensive synthetic dataset simulating common dermoscopic degradations, including Gaussian noise, variations in brightness and contrast, and blur. Extensive experiments were conducted, evaluating our proposed method against several established baseline models and analyzing the impact of various loss functions and optimizers to determine the optimal configuration. The results show that EnhanceNet-U consistently outperformed all competing models, demonstrating a peak improvement of 15.75% in SSIM and 15% in PSNR when compared to the lowest-performing DnCNN model. The combination of perceptual loss and MSE emerged as the most effective loss function for balancing quantitative accuracy with perceptual quality. These findings validate our proposed method, proving its capability to restore high-quality images under realistic degradation scenarios and highlighting its potential as a robust solution for the complexities of dermoscopic image enhancement.</div></div>","PeriodicalId":100554,"journal":{"name":"Franklin Open","volume":"13 ","pages":"Article 100405"},"PeriodicalIF":0.0,"publicationDate":"2025-10-21","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"145341039","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

{"title":"PhishHunter-XLD: An ensemble approach integrating machine learning and deep learning for phishing URL classification","authors":"Tirth Doshi ,&nbsp;Vishva Patel ,&nbsp;Nemil Shah ,&nbsp;Debabrata Swain ,&nbsp;Debabala Swain ,&nbsp;Biswaranjan Acharya","doi":"10.1016/j.fraope.2025.100349","DOIUrl":"10.1016/j.fraope.2025.100349","url":null,"abstract":"<div><div>Phishing continues to pose a significant cybersecurity threat by deceiving users into disclosing sensitive information through maliciously crafted URLs. Traditional detection methods, including blacklists and heuristic analyses, have proven inadequate against evolving phishing techniques due to their reliance on static patterns and manual updates. In this study, a weighted voting ensemble framework has been proposed, integrating semantic feature extraction using DistilBERT with classical machine learning classifiers (XGBoost) and deep learning models (LSTM) to enhance phishing URL detection. Model complementarity has been leveraged XGBoost captures explicit lexical features, LSTM models sequential dependencies, and DistilBERT extracts contextual semantics resulting in an adaptive decision boundary that improves generalization and reduces false positives. Extensive experiments conducted on large-scale benchmark datasets, such as the “Phishing Site URLs” and “Malicious URLs” datasets, have demonstrated that the proposed ensemble framework achieves a detection accuracy of 99.83% with low computational latency. Furthermore, the system has been deployed via Streamlit, providing a real time, interactive interface for cybersecurity practitioners. Future work will explore optimization strategies, including model pruning, quantization, and adversarial training, to further enhance efficiency, scalability, and resilience against emerging zero-day phishing techniques.</div></div>","PeriodicalId":100554,"journal":{"name":"Franklin Open","volume":"12 ","pages":"Article 100349"},"PeriodicalIF":0.0,"publicationDate":"2025-09-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"145219141","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

{"title":"Image feature extraction techniques: A comprehensive review","authors":"Sudhakar Hallur ,&nbsp;Anil Gavade","doi":"10.1016/j.fraope.2025.100366","DOIUrl":"10.1016/j.fraope.2025.100366","url":null,"abstract":"<div><div>This comprehensive review explores the landscape of image feature extraction techniques, which form the cornerstone of modern image processing and computer vision applications. Feature extraction serves the critical function of transforming raw image data into informative and compact representations, enabling efficient analysis, recognition, and classification. The paper systematically categorizes and analyzes methods based on geometric, statistical, texture, color, and conceptual features. Geometric features capture structural relationships and object shapes, while statistical features provide quantitative descriptors of intensity distributions. Texture-based techniques such as Local Binary Patterns (LBP) and Gray Level Co-occurrence Matrix (GLCM) highlight surface characteristics and spatial patterns. Color features, including histograms and moments, model chromatic information vital for retrieval and segmentation tasks. The review also discusses the emerging role of deep learning in extracting hierarchical and abstract features, which offer superior adaptability and semantic richness. For each category, the strengths, limitations, computational efficiency, and domain-specific applicability are critically evaluated. The paper concludes by emphasizing the merits of multi-feature fusion approaches that integrate diverse descriptors to enhance robustness and accuracy in image understanding tasks. This survey aims to guide future research by offering a foundational and comparative perspective on classical and contemporary feature extraction strategies.</div></div>","PeriodicalId":100554,"journal":{"name":"Franklin Open","volume":"12 ","pages":"Article 100366"},"PeriodicalIF":0.0,"publicationDate":"2025-09-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"145157110","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

{"title":"An optimized hybrid block technique framework for partial differential equations by exploring radial basis function neural networks","authors":"Sabastine Emmanuel ,&nbsp;Saratha Sathasivam ,&nbsp;Muideen O. Ogunniran ,&nbsp;Mustafa Bayram","doi":"10.1016/j.fraope.2025.100332","DOIUrl":"10.1016/j.fraope.2025.100332","url":null,"abstract":"<div><div>This article explores recent advances in integrating artificial neural networks (ANNs) with numerical methods, focusing on an optimized two-stage hybrid block method for solving complex partial differential equations (PDEs). The method combines ANNs with traditional solvers to improve computational efficiency and accuracy, using a Radial Basis Function Neural Network (RBFNN) for optimization. This approach enhances stability, convergence, and adaptability, especially in multidimensional problems, and reduces computational costs while maintaining precision in fields like fluid dynamics and electromagnetism. Simulations show the method outperforms conventional solvers, enabling more efficient real-time simulations in engineering and mathematics.</div></div>","PeriodicalId":100554,"journal":{"name":"Franklin Open","volume":"12 ","pages":"Article 100332"},"PeriodicalIF":0.0,"publicationDate":"2025-09-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"144932022","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

{"title":"A comparative review of radial and axial Flux PMSMs: Innovations in topology, design, and control","authors":"Preet Samanta,&nbsp;Sagar Babu Mitikiri,&nbsp;Pranay Krishna Sahay,&nbsp;Vedantham Lakshmi Srinivas","doi":"10.1016/j.fraope.2025.100341","DOIUrl":"10.1016/j.fraope.2025.100341","url":null,"abstract":"<div><div>Permanent Magnet Synchronous Motors (PMSMs) are widely adopted in electric propulsion systems due to their compactness, superior torque density, and high efficiency. This review provides a comprehensive comparative analysis of Radial Flux PMSMs (RFPMSMs and Axial Flux PMSMs (AFPMSMs, emphasizing recent advancements in topology, design optimization, and control strategies. Specifically, it highlights innovations such as spoke-type and V-shaped rotor designs Halbach magnet arrays, and multi-rotor axial flux configurations, which have demonstrated up to 28% improvement in torque density and reduction in cogging torque by 65% through optimized winding and magnet layouts. In terms of design methodology, the use of metaheuristic algorithms – including Genetic Algorithms (GA), Particle Swarm Optimization (PSO), and Artificial Neural Networks (ANNs) – has shown to enhance electromagnetic and thermal performance while improving overall efficiency by as much as 15% under real driving cycles. Additionally, this review explores advanced control strategies, such as Model Predictive Control (MPC), Sliding Mode Control (SMC), and sensorless vector control, which have contributed to a 20%–30% improvement in dynamic response and reduced total harmonic distortion in torque output by up to 40%. This study not only identifies the technological trends that are shaping the next generation of high-performance PMSMs but also critically outlines current limitations and future research opportunities in the domain.</div></div>","PeriodicalId":100554,"journal":{"name":"Franklin Open","volume":"12 ","pages":"Article 100341"},"PeriodicalIF":0.0,"publicationDate":"2025-09-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"145057082","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

{"title":"Power flow management in hybrid AC/DC microgrids using the Artificial Bee Colony metaheuristic algorithm: A comparative study","authors":"Ssadik Charadi ,&nbsp;Houssame Eddine Chakir ,&nbsp;Abdelbari Redouane ,&nbsp;Youssef Akarne ,&nbsp;Abdennebi El Hasnaoui ,&nbsp;Mehdi Et-taoussi","doi":"10.1016/j.fraope.2025.100356","DOIUrl":"10.1016/j.fraope.2025.100356","url":null,"abstract":"<div><div>This paper presents an advanced power flow management strategy for a hybrid AC/DC microgrid integrating renewable energy sources such as photovoltaic systems and wind turbines, alongside an energy storage system and a gas turbine as a polluting source. The microgrid is interconnected with the main grid, enabling bidirectional energy exchanges to efficiently balance supply and demand. The main technical innovation is the development of an economic dispatch method based on the Artificial Bee Colony metaheuristic algorithm, which simultaneously optimizes both active and reactive power flows using discrete-time control techniques. The approach features a unified objective function that combines cost minimization and reduction of greenhouse gas emissions, enabling coordinated decision making for improved economic efficiency and environmental sustainability. A comparative analysis with the Particle Swarm Optimization algorithm and a conventional rule-based approach demonstrates the superior performance of Artificial Bee Colony in several key aspects. Specifically, Artificial Bee Colony achieves a 43.1% reduction in <span><math><mrow><mi>C</mi><msub><mrow><mi>O</mi></mrow><mrow><mn>2</mn></mrow></msub></mrow></math></span> emissions compared to the conventional approach and 33.9% compared to Particle Swarm Optimization. Additionally, it lowers operational costs by 27.54% relative to the conventional method, outperforming Particle Swarm Optimization, which achieves a 14.83% reduction. Furthermore, Artificial Bee Colony increases the utilization of the Energy Storage System, reaching 463.05 kVA, compared to 284.07 kVA with Particle Swarm Optimization and 153.85 kVA with the conventional approach. The algorithm also reduces dependency on the main grid by 29. 30%, surpassing particle swarm optimization (12. 29%) and improves the integration of renewable energy sources by 12. 94%, compared to 5. 40% for particle swarm optimization. The simulation was conducted using MATLAB® 2024a, leveraging 24-hour ahead meteorological forecasts and load profiles to ensure optimal and adaptive operation under dynamic conditions. These results highlight the effectiveness of Artificial Bee Colony in addressing the challenges of hybrid microgrid management, making it a robust and scalable solution for modern energy systems.</div></div>","PeriodicalId":100554,"journal":{"name":"Franklin Open","volume":"12 ","pages":"Article 100356"},"PeriodicalIF":0.0,"publicationDate":"2025-09-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"145057083","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

{"title":"Novel coprime MIMO configurations for DOA estimation with increased degrees of freedom","authors":"Rajen Kumar Patra ,&nbsp;Anindya Sundar Dhar","doi":"10.1016/j.fraope.2025.100374","DOIUrl":"10.1016/j.fraope.2025.100374","url":null,"abstract":"<div><div>In this work, we propose some coprime multiple input multiple output (MIMO) configurations for direction-of-arrival (DOA) estimation, which achieve more degrees of freedom (DOF) than all the existing coprime MIMO configurations. We know that in a coprime MIMO configuration, we can exploit the difference coarray of the sum coarray (DCSC) by vectorizing the covariance matrix. The transmitter array of each of the proposed configurations uses a coprime array, where one of the subarrays of the array is shifted. In the first proposed configuration, the inter-element spacing of the receiver array is increased by an expansion factor, and it is shown that the configuration achieves a significant number of uniform and unique DOF. In the second configuration, we increase the expansion factor to further enhance the uniform and unique DOF. The expressions of the number of uniform and unique DOF are provided for both the proposed configurations. All the required simulations are carried out to evaluate the performance of the proposed configurations and show their advantages.</div></div>","PeriodicalId":100554,"journal":{"name":"Franklin Open","volume":"12 ","pages":"Article 100374"},"PeriodicalIF":0.0,"publicationDate":"2025-09-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"145219143","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

{"title":"CAChaIoT: Hybrid lightweight image encryption for IoT using cellular automata and chaotic maps","authors":"Biswarup Yogi ,&nbsp;Ajoy Kumar Khan","doi":"10.1016/j.fraope.2025.100361","DOIUrl":"10.1016/j.fraope.2025.100361","url":null,"abstract":"<div><div>With the rise of the Internet of Things (IoT), the protection of image data is a challenging task due to the limited resources of the devices and the ability to process data in an efficient and timely manner. The proposed lightweight image encryption scheme uses Cellular Automata Rule 30 and the Tinkerbell map within the same framework to deliver both adequate data protection and computational efficiency. The work begins by dividing the image into its red, green, and blue RGB channels, starting with Tinkerbell map-based encryption, followed by Cellular Automata Rule 30-based encryption for improved diffusion. We implemented the method and achieved a very high level of encryption, as evidenced by an NPCR (Number of Pixels Change Rate) of 99.6261, which assures strong security.. The proposed algorithm achieves strong security metrics, including a UACI of 49.86%, which exceeds the baseline of 33.33% and indicates high sensitivity to input changes, making it robust against differential attacks. This method is ideal in IoT devices because it is efficient and simple. Future studies can explore the inclusion of other cryptographic methods, as well as optimisation on different IoT platforms to achieve better scalability and security.</div></div>","PeriodicalId":100554,"journal":{"name":"Franklin Open","volume":"12 ","pages":"Article 100361"},"PeriodicalIF":0.0,"publicationDate":"2025-09-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"145104527","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

{"title":"Artificial neural network-based prediction of tribological performance of coatings fabricated by thermal spray techniques for nuclear energy applications","authors":"Shubhangi Suryawanshi ,&nbsp;Digvijay G. Bhosale ,&nbsp;Hitesh Vasudev ,&nbsp;Sanjay Rukhande ,&nbsp;Vikas Panwar","doi":"10.1016/j.fraope.2025.100364","DOIUrl":"10.1016/j.fraope.2025.100364","url":null,"abstract":"<div><div>Nuclear energy systems are subject to harsh operating conditions, such as wide temperature variations and humidity changes with altitude shifts, which can lead to wear in different devices. The coatings of single or multi-layer are excellent solution for control over the wear rate. The beneficial characteristics of these multilayer coatings may be expertly customised for any application using artificial neural network (ANN) based models. Accordingly present work is for the rate of sliding wear and friction of hard coatings sprayed by two various types of thermal spray processes is predicted using a model created using an artificial neural network approach. The mechanical and tribological performances of four types of deposits, including coatings made of WC20Cr₃C₂7Ni and NiCrSiBFe that were deposited using APS and HVOF, were examined and analysed. The controllable factors considered in the developed ANN model are sliding distance, load, velocity, temperature of tribo-chamber, standoff distance, gun transverse speed, powder feed rate, porosity, and nanohardness. The ANN made it feasible to study the interactions between the two output parameters and the nine input parameters. The optimal outcome of deposited film characteristics like friction and wear are evaluated using regression and performance curve analysis. The studied data demonstrates an excellent consistency between the experimental and predicted tribological parameters. The evaluation of the mean impact value was employed to quantitatively ascertain the significance of each input parameter, with the objective of enhancing prediction accuracy.</div></div>","PeriodicalId":100554,"journal":{"name":"Franklin Open","volume":"12 ","pages":"Article 100364"},"PeriodicalIF":0.0,"publicationDate":"2025-09-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"145121019","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

{"title":"ReLiris: Lung damage pre-screening via iridology","authors":"Trong Thua Huynh ,&nbsp;De Thu Huynh ,&nbsp;Hoang Phat Vu ,&nbsp;Minh Tuan Nguyen ,&nbsp;Thi Tuyet Hai Nguyen ,&nbsp;Lam Thanh Tu","doi":"10.1016/j.fraope.2025.100362","DOIUrl":"10.1016/j.fraope.2025.100362","url":null,"abstract":"<div><div>Spotting lung damage sooner can make a real difference in patient recovery by opening the door to quicker treatment. Yet, conventional exams like CT scans or X-rays tend to be costly, slow, and demand bulky, specialized machinery. In this study, we propose ReLiris, a hybrid deep learning framework designed to spot early signs of lung damage by analyzing simple images of the human iris. At its core, ReLiris uses a ResNet-based convolutional network to pull out meaningful image features, then hands those over to a pair of bidirectional LSTM layers equipped with a self-attention mechanism. This combination helps the model focus on the most important patterns and prevents issues like inactive neurons. Both the ResNet and LSTM parameters are updated in tandem during a single training pass. The model’s final output, a probability score for lung damage, is optimized to minimize its discrepancy from the true labels. We also employ the Swish activation and Focal Loss to further boost learning of complex features. Trained on a set of 10,866 labeled iris images, ReLiris learns to distinguish between healthy and at-risk lungs. In our experiments, it achieved 96.29% accuracy – outperforming standalone ResNet and several well-known benchmarks – with sensitivity above 96%, inference time around 6 s and a ROC AUC reach to 0.9923. These results suggest that simple iris scans, when paired with the right deep learning tools, could become an affordable, noninvasive way to detect lung damage at its earliest stages, potentially improving patient outcomes and survival rates.</div></div>","PeriodicalId":100554,"journal":{"name":"Franklin Open","volume":"12 ","pages":"Article 100362"},"PeriodicalIF":0.0,"publicationDate":"2025-09-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"145219140","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}