alexandria engineering journal最新文献_第8页

Nanotechnology-driven targeted therapies for gastrointestinal diseases: State-of-the-art strategies and future directions 纳米技术驱动的胃肠道疾病靶向治疗：最先进的策略和未来方向

IF 6.8 2区工程技术 Q1 ENGINEERING, MULTIDISCIPLINARY

alexandria engineering journal

Pub Date : 2026-01-01 Epub Date: 2026-01-07 DOI: 10.1016/j.aej.2025.12.064

Dong Li , An Yan , Jiexin Chen , Mengmeng Sun , Min He , Wanmin Liu

Recent advances in nanotechnology have enabled transformative strategies for addressing the multifaceted challenges associated with gastrointestinal (GI) diseases. These disorders, encompassing inflammatory bowel diseases, colorectal cancer, and infections like Helicobacter pylori, often demand precise therapeutic delivery and reduced systemic toxicity—requirements that conventional treatments struggle to meet. Nanomedicine offers a paradigm shift by leveraging diverse nanocarriers such as liposomes, polymeric nanoparticles, dendrimers, exosomes, and inorganic platforms to enhance drug solubility, protect labile drugs, and achieve controlled and site-specific delivery. This review outlines the latest passive and active targeting strategies, including ligand-mediated approaches and stimuli-responsive release systems, which improve drug accumulation at disease sites. It also highlights oral delivery challenges posed by the gastrointestinal microenvironment and discusses engineering solutions such as surface modification and protective coatings. Furthermore, plant-derived and cell-derived vesicle-like nanoparticles are emerging as bioinspired delivery vectors offering biocompatibility and immune modulation. We synthesize current preclinical and clinical progress, emphasizing nanomedicines already under investigation or approved for GI conditions. Despite promising data, barriers such as scalable production, regulatory complexities, and long-term safety concerns remain. This review concludes by charting future directions that focus on personalized therapy, biosensing integration, and intelligent nanoplatforms. By dissecting the intersection of materials science, pharmacology, and gastroenterology, this work provides a roadmap for accelerating translational breakthroughs in GI-targeted nanotherapies.

纳米技术的最新进展使解决与胃肠道疾病相关的多方面挑战的变革性战略成为可能。这些疾病，包括炎症性肠病、结肠直肠癌和幽门螺杆菌感染，通常需要精确的治疗递送和降低全身毒性，这是传统治疗难以满足的要求。纳米医学通过利用各种纳米载体（如脂质体、聚合纳米颗粒、树突分子、外泌体和无机平台）来提高药物的溶解度，保护不稳定的药物，并实现控制和位点特异性递送，从而提供了一种范式转变。本文概述了最新的被动和主动靶向策略，包括配体介导的方法和刺激反应性释放系统，它们可以改善疾病部位的药物积累。它还强调了胃肠道微环境带来的口服递送挑战，并讨论了表面改性和保护涂层等工程解决方案。此外，植物源性和细胞源性囊泡样纳米颗粒正在成为生物启发的递送载体，具有生物相容性和免疫调节功能。我们综合当前的临床前和临床进展，强调已经在研究或批准用于胃肠道疾病的纳米药物。尽管数据很有前景，但可扩展生产、监管复杂性和长期安全问题等障碍仍然存在。本文最后展望了个性化治疗、生物传感集成和智能纳米平台的未来发展方向。通过剖析材料科学、药理学和胃肠病学的交叉，这项工作为加速gi靶向纳米治疗的转化突破提供了路线图。

{"title":"Nanotechnology-driven targeted therapies for gastrointestinal diseases: State-of-the-art strategies and future directions","authors":"Dong Li , An Yan , Jiexin Chen , Mengmeng Sun , Min He , Wanmin Liu","doi":"10.1016/j.aej.2025.12.064","DOIUrl":"10.1016/j.aej.2025.12.064","url":null,"abstract":"<div><div>Recent advances in nanotechnology have enabled transformative strategies for addressing the multifaceted challenges associated with gastrointestinal (GI) diseases. These disorders, encompassing inflammatory bowel diseases, colorectal cancer, and infections like <em>Helicobacter pylori</em>, often demand precise therapeutic delivery and reduced systemic toxicity—requirements that conventional treatments struggle to meet. Nanomedicine offers a paradigm shift by leveraging diverse nanocarriers such as liposomes, polymeric nanoparticles, dendrimers, exosomes, and inorganic platforms to enhance drug solubility, protect labile drugs, and achieve controlled and site-specific delivery. This review outlines the latest passive and active targeting strategies, including ligand-mediated approaches and stimuli-responsive release systems, which improve drug accumulation at disease sites. It also highlights oral delivery challenges posed by the gastrointestinal microenvironment and discusses engineering solutions such as surface modification and protective coatings. Furthermore, plant-derived and cell-derived vesicle-like nanoparticles are emerging as bioinspired delivery vectors offering biocompatibility and immune modulation. We synthesize current preclinical and clinical progress, emphasizing nanomedicines already under investigation or approved for GI conditions. Despite promising data, barriers such as scalable production, regulatory complexities, and long-term safety concerns remain. This review concludes by charting future directions that focus on personalized therapy, biosensing integration, and intelligent nanoplatforms. By dissecting the intersection of materials science, pharmacology, and gastroenterology, this work provides a roadmap for accelerating translational breakthroughs in GI-targeted nanotherapies.</div></div>","PeriodicalId":7484,"journal":{"name":"alexandria engineering journal","volume":"135 ","pages":"Pages 209-234"},"PeriodicalIF":6.8,"publicationDate":"2026-01-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"145923618","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":2,"RegionCategory":"工程技术","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

引用次数: 0

Hybrid neurosymbolic causal inference Q-network (NS-CIQN) for optimizing energy efficiency in consumer electronics 用于优化消费电子产品能源效率的混合神经符号因果推理q网络（NS-CIQN）

IF 6.8 2区工程技术 Q1 ENGINEERING, MULTIDISCIPLINARY

alexandria engineering journal

Pub Date : 2026-01-01 Epub Date: 2026-01-07 DOI: 10.1016/j.aej.2025.12.052

Jian Chen , Shaowen Ma , Tingting Yang , Bo Qiu

The rapid growth of consumer electronics, such as smartphones, laptops, smart televisions, and home appliances, now accounts for a significant share of global electricity use and carbon emissions. Although most artificial intelligence (AI) energy-saving methods perform well, their limited interpretability often reduces trust among users and manufacturers. This study presents Neurosymbolic Causal Inference Q-Network (NS-CIQN), a hybrid Neurosymbolic framework that combines effectiveness with explainability. The framework uses causal graph neural networks to identify true cause-and-effect relationships, such as occupancy, time of day, temperature, and charging behavior. It also integrates symbolic reasoning rules that are both accessible and verifiable, along with a Deep Q-Network that learns optimal power-management strategies through reinforcement learning. A modified Whale Optimization algorithm fine-tunes solutions while preserving user comfort. Testing on the UK-DALE (UK Domestic Appliance-Level Electricity Dataset) and Reference Energy Disaggregation Dataset (REDD) household datasets, updated with 2025 regional carbon-intensity data, shows that NS-CIQN achieves an 88.4 % F1-score in recognizing appliance usage patterns, outperforming standard deep reinforcement learning and Random Forest methods, which score 81–83 %. NS-CIQN also reduces total energy consumption by 14.8–17.3 % in realistic simulations without affecting user comfort. The system’s causal graphs offer transparent, actionable explanations to support device manufacturers, smart-home platforms, and policymakers in advancing sustainability

智能手机、笔记本电脑、智能电视和家用电器等消费电子产品的快速增长，目前在全球用电量和碳排放中占很大份额。尽管大多数人工智能（AI）节能方法表现良好，但其有限的可解释性往往会降低用户和制造商之间的信任。本研究提出了神经符号因果推理q网络（NS-CIQN），这是一个结合了有效性和可解释性的混合神经符号框架。该框架使用因果图神经网络来识别真正的因果关系，例如占用、一天中的时间、温度和充电行为。它还集成了可访问和可验证的符号推理规则，以及通过强化学习学习最佳电源管理策略的Deep Q-Network。改进的鲸鱼优化算法微调解决方案，同时保持用户的舒适性。在UK- dale（英国家用电器级电力数据集）和参考能源分解数据集（REDD）家庭数据集上的测试，更新了2025年的区域碳强度数据，表明NS-CIQN在识别家电使用模式方面达到了88.4 % f1得分，优于标准的深度强化学习和随机森林方法，得分为81-83 %。在不影响用户舒适度的情况下，NS-CIQN还在现实模拟中减少了14.8-17.3 %的总能耗。该系统的因果图提供了透明、可操作的解释，以支持设备制造商、智能家居平台和政策制定者推进可持续性

{"title":"Hybrid neurosymbolic causal inference Q-network (NS-CIQN) for optimizing energy efficiency in consumer electronics","authors":"Jian Chen , Shaowen Ma , Tingting Yang , Bo Qiu","doi":"10.1016/j.aej.2025.12.052","DOIUrl":"10.1016/j.aej.2025.12.052","url":null,"abstract":"<div><div>The rapid growth of consumer electronics, such as smartphones, laptops, smart televisions, and home appliances, now accounts for a significant share of global electricity use and carbon emissions. Although most artificial intelligence (AI) energy-saving methods perform well, their limited interpretability often reduces trust among users and manufacturers. This study presents Neurosymbolic Causal Inference Q-Network (NS-CIQN), a hybrid Neurosymbolic framework that combines effectiveness with explainability. The framework uses causal graph neural networks to identify true cause-and-effect relationships, such as occupancy, time of day, temperature, and charging behavior. It also integrates symbolic reasoning rules that are both accessible and verifiable, along with a Deep Q-Network that learns optimal power-management strategies through reinforcement learning. A modified Whale Optimization algorithm fine-tunes solutions while preserving user comfort. Testing on the UK-DALE (UK Domestic Appliance-Level Electricity Dataset) and Reference Energy Disaggregation Dataset (REDD) household datasets, updated with 2025 regional carbon-intensity data, shows that NS-CIQN achieves an 88.4 % F1-score in recognizing appliance usage patterns, outperforming standard deep reinforcement learning and Random Forest methods, which score 81–83 %. NS-CIQN also reduces total energy consumption by 14.8–17.3 % in realistic simulations without affecting user comfort. The system’s causal graphs offer transparent, actionable explanations to support device manufacturers, smart-home platforms, and policymakers in advancing sustainability</div></div>","PeriodicalId":7484,"journal":{"name":"alexandria engineering journal","volume":"135 ","pages":"Pages 327-341"},"PeriodicalIF":6.8,"publicationDate":"2026-01-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"145923610","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":2,"RegionCategory":"工程技术","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

引用次数: 0

Graph-based spatial-temporal networks for traffic speed prediction in intelligent transport systems 智能交通系统中基于图的交通速度预测时空网络

IF 6.8 2区工程技术 Q1 ENGINEERING, MULTIDISCIPLINARY

alexandria engineering journal

Pub Date : 2026-01-01 Epub Date: 2026-01-08 DOI: 10.1016/j.aej.2025.12.055

Harishankar K. Nair , Logesh Ravi , Malathi Devarajan , P. Saravanan

Traffic speed forecasting is among the key challenges in improving traffic management and urban mobility in the framework of the Intelligent Transport Systems. The complexity and nonlinear nature of traffic data coupled with learning dynamic spatial-temporal dependencies present a significant challenge that the traditional forecasting methods are unable to address, due to their inability to model dynamic changes in the traffic network. Our paper presents two new deep learning-based models, ST-GIncep (Spatial-Temporal Gated Inception Network) as well as ST-GTransLSTM (Spatial-Temporal Graph Transformer LSTM). The ST-Incep uses a multi-scale convolution architecture along with gating mechanisms that are able to capture spatial features at various scales along with Matrix LSTM to model temporal features. The ST-GTransLSTM combines the ability to the graph transformer to capture spatial information efficiently coupled with the capacity of the LSTM to process sequences data to create a hybrid Encoder-Decoder architecture allowing effective modeling long-term temporal dependencies. The frameworks presented in this paper are tested on the real-traffic dataset PemSD7. The experimental results depict both proposed models outperform existing baseline frameworks with respect to predictive accuracy and robustness. The performance of the approaches presented on noisy data and efficiency on low-traffic networks demonstrate their utility in real-world ITS applications.

交通速度预测是在智能交通系统框架下改善交通管理和城市机动性的主要挑战之一。交通数据的复杂性和非线性，加上学习动态时空依赖关系，对传统的预测方法提出了重大挑战，因为它们无法模拟交通网络的动态变化。本文提出了两个新的基于深度学习的模型，ST-GIncep（时空门通初始网络）和ST-GTransLSTM（时空图转换LSTM）。ST-Incep使用多尺度卷积架构和门控机制，能够捕获各种尺度的空间特征，以及矩阵LSTM来模拟时间特征。ST-GTransLSTM结合了图形转换器有效捕获空间信息的能力，以及LSTM处理序列数据的能力，创建了一个混合编码器-解码器架构，允许有效地建模长期时间依赖性。本文提出的框架在实际交通数据集PemSD7上进行了测试。实验结果表明，所提出的模型在预测精度和鲁棒性方面优于现有的基线框架。所提出的方法在噪声数据上的性能和在低流量网络上的效率证明了它们在现实世界ITS应用中的实用性。

{"title":"Graph-based spatial-temporal networks for traffic speed prediction in intelligent transport systems","authors":"Harishankar K. Nair , Logesh Ravi , Malathi Devarajan , P. Saravanan","doi":"10.1016/j.aej.2025.12.055","DOIUrl":"10.1016/j.aej.2025.12.055","url":null,"abstract":"<div><div>Traffic speed forecasting is among the key challenges in improving traffic management and urban mobility in the framework of the Intelligent Transport Systems. The complexity and nonlinear nature of traffic data coupled with learning dynamic spatial-temporal dependencies present a significant challenge that the traditional forecasting methods are unable to address, due to their inability to model dynamic changes in the traffic network. Our paper presents two new deep learning-based models, ST-GIncep (Spatial-Temporal Gated Inception Network) as well as ST-GTransLSTM (Spatial-Temporal Graph Transformer LSTM). The ST-Incep uses a multi-scale convolution architecture along with gating mechanisms that are able to capture spatial features at various scales along with Matrix LSTM to model temporal features. The ST-GTransLSTM combines the ability to the graph transformer to capture spatial information efficiently coupled with the capacity of the LSTM to process sequences data to create a hybrid Encoder-Decoder architecture allowing effective modeling long-term temporal dependencies. The frameworks presented in this paper are tested on the real-traffic dataset PemSD7. The experimental results depict both proposed models outperform existing baseline frameworks with respect to predictive accuracy and robustness. The performance of the approaches presented on noisy data and efficiency on low-traffic networks demonstrate their utility in real-world ITS applications.</div></div>","PeriodicalId":7484,"journal":{"name":"alexandria engineering journal","volume":"135 ","pages":"Pages 384-399"},"PeriodicalIF":6.8,"publicationDate":"2026-01-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"145923607","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":2,"RegionCategory":"工程技术","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

引用次数: 0

Invasive weed optimization based metaheuristic approach for solving constrained risk budgeted portfolio selection problem 基于入侵杂草优化的约束风险投资组合选择元启发式方法

IF 6.8 2区工程技术 Q1 ENGINEERING, MULTIDISCIPLINARY

alexandria engineering journal

Pub Date : 2026-01-01 Epub Date: 2026-01-07 DOI: 10.1016/j.aej.2025.12.067

Zubair Ashraf , Mohammad Shahid , Lamaan Sami , Faraz Hasan , Mohd Shamim

Portfolio management focuses on investing in the financial sector to achieve the highest return while tolerating the lowest risk. The optimal financial allocation has long been considered one of the essential aspects of risk-adjusted financial sector investment. Therefore, many optimization techniques have been established to maximize the return on risk. This paper presents a novel framework named the risk-budgeted portfolio selection (RBPS) model, which allocates the total risk of a portfolio across different securities by incorporating risk budgeting (RB) levels to ensure the portfolio's risk is diversified while maximizing the Sharpe ratio. To address the proposed RBPS model, an invasive weed optimization (IWO) algorithm-based solution method is suggested, and risk budgeting constraints are accommodated using resilient and flexible repairing procedures. Experiments have been performed using two newly created datasets from the Sensex of the Bombay Stock Exchange and the National Stock Exchange from India. The percentage improvement of the maximum Sharpe ratio obtained by IWO is up to 1.95 % at RB% = 12.5 among its peer's algorithms. Moreover, the experiments have been extended to global benchmark datasets to evaluate the proposed approach. Finally, statistical analysis is conducted to test the significance of improvement in the RBPS model.

投资组合管理的重点是投资于金融部门，以实现最高的回报，同时承受最低的风险。长期以来，最优金融配置一直被认为是风险调整金融部门投资的重要方面之一。因此，建立了许多优化技术来最大化风险回报。本文提出了风险预算投资组合选择（RBPS）模型，该模型通过纳入风险预算（RB）水平来分配投资组合的总风险，以确保投资组合的风险分散，同时最大化夏普比率。针对RBPS模型，提出了一种基于入侵杂草优化（IWO）算法的求解方法，并采用弹性和柔性修复程序来适应风险预算约束。实验使用了两个新创建的数据集，这些数据集来自孟买证券交易所的Sensex和印度国家证券交易所。在RB% = 12.5时，IWO算法获得的最大夏普比在同类算法中提高了1.95 %。此外，实验已经扩展到全球基准数据集来评估所提出的方法。最后进行统计分析，检验RBPS模型改进的显著性。

{"title":"Invasive weed optimization based metaheuristic approach for solving constrained risk budgeted portfolio selection problem","authors":"Zubair Ashraf , Mohammad Shahid , Lamaan Sami , Faraz Hasan , Mohd Shamim","doi":"10.1016/j.aej.2025.12.067","DOIUrl":"10.1016/j.aej.2025.12.067","url":null,"abstract":"<div><div>Portfolio management focuses on investing in the financial sector to achieve the highest return while tolerating the lowest risk. The optimal financial allocation has long been considered one of the essential aspects of risk-adjusted financial sector investment. Therefore, many optimization techniques have been established to maximize the return on risk. This paper presents a novel framework named the risk-budgeted portfolio selection (RBPS) model, which allocates the total risk of a portfolio across different securities by incorporating risk budgeting (RB) levels to ensure the portfolio's risk is diversified while maximizing the Sharpe ratio. To address the proposed RBPS model, an invasive weed optimization (IWO) algorithm-based solution method is suggested, and risk budgeting constraints are accommodated using resilient and flexible repairing procedures. Experiments have been performed using two newly created datasets from the Sensex of the Bombay Stock Exchange and the National Stock Exchange from India. The percentage improvement of the maximum Sharpe ratio obtained by IWO is up to 1.95 % at RB% = 12.5 among its peer's algorithms. Moreover, the experiments have been extended to global benchmark datasets to evaluate the proposed approach. Finally, statistical analysis is conducted to test the significance of improvement in the RBPS model.</div></div>","PeriodicalId":7484,"journal":{"name":"alexandria engineering journal","volume":"135 ","pages":"Pages 273-317"},"PeriodicalIF":6.8,"publicationDate":"2026-01-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"145923681","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":2,"RegionCategory":"工程技术","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

引用次数: 0

Study on the tribological properties of AlN/AlCrN composite coatings on the surface of TA15 alloy TA15合金表面AlN/AlCrN复合涂层摩擦学性能研究

IF 6.8 2区工程技术 Q1 ENGINEERING, MULTIDISCIPLINARY

alexandria engineering journal

Pub Date : 2026-01-01 Epub Date: 2026-01-07 DOI: 10.1016/j.aej.2025.12.063

Guowei Mo, Hui Yu

To address the dual challenges of poor adhesion and insufficient high-temperature wear resistance of protective coatings on TA15 titanium alloy, this study proposes a tailored AlN/AlCrN composite coating design. The coatings were fabricated via reactive magnetron sputtering with an optimized AlN transition layer. This architecture resulted in a dense columnar structure, significantly enhancing the surface properties. The composite coating achieved an average hardness of 29.87 GPa—three times that of the substrate—and an interfacial bonding strength of 44.17 ± 1.85 MPa, a 16.6 % increase over a single AlCrN layer. More importantly, the coating exhibited exceptional tribological stability across a wide temperature range. Under a 300 g load at room temperature, the friction coefficient dropped to 0.28 (46 % lower than the substrate), and the wear rate decreased by 73 %. Remarkably, at 600°C, the friction coefficient further reduced to 0.21, with the wear rate being only one-sixth of the substrate, which is attributed to the in-situ formation of a lubricious Al/Cr-rich oxide layer. This work provides an effective surface modification strategy, broadening the application horizon of TA15 alloy under extreme conditions.

针对TA15钛合金防护涂层附着力差和高温耐磨性不足的双重挑战，本研究提出了定制化AlN/AlCrN复合涂层设计。采用反应磁控溅射法制备了优化后的氮化铝过渡层。这种结构形成了致密的柱状结构，显著提高了表面性能。复合涂层的平均硬度为29.87 gpa，是基体的3倍，界面结合强度为44.17 ± 1.85 MPa，比单一AlCrN层提高了16.6 %。更重要的是，涂层在很宽的温度范围内表现出优异的摩擦学稳定性。在300 g的室温载荷下，摩擦系数降至0.28（比基体低46 %），磨损率下降73 %。值得注意的是，在600°C时，摩擦系数进一步降低到0.21，磨损率仅为基体的六分之一，这是由于原位形成了富含Al/ cr的着色氧化层。本研究为TA15合金在极端条件下的表面改性提供了一种有效的策略，拓宽了TA15合金在极端条件下的应用前景。

{"title":"Study on the tribological properties of AlN/AlCrN composite coatings on the surface of TA15 alloy","authors":"Guowei Mo, Hui Yu","doi":"10.1016/j.aej.2025.12.063","DOIUrl":"10.1016/j.aej.2025.12.063","url":null,"abstract":"<div><div>To address the dual challenges of poor adhesion and insufficient high-temperature wear resistance of protective coatings on TA15 titanium alloy, this study proposes a tailored AlN/AlCrN composite coating design. The coatings were fabricated via reactive magnetron sputtering with an optimized AlN transition layer. This architecture resulted in a dense columnar structure, significantly enhancing the surface properties. The composite coating achieved an average hardness of 29.87 GPa—three times that of the substrate—and an interfacial bonding strength of 44.17 ± 1.85 MPa, a 16.6 % increase over a single AlCrN layer. More importantly, the coating exhibited exceptional tribological stability across a wide temperature range. Under a 300 g load at room temperature, the friction coefficient dropped to 0.28 (46 % lower than the substrate), and the wear rate decreased by 73 %. Remarkably, at 600°C, the friction coefficient further reduced to 0.21, with the wear rate being only one-sixth of the substrate, which is attributed to the in-situ formation of a lubricious Al/Cr-rich oxide layer. This work provides an effective surface modification strategy, broadening the application horizon of TA15 alloy under extreme conditions.</div></div>","PeriodicalId":7484,"journal":{"name":"alexandria engineering journal","volume":"135 ","pages":"Pages 262-272"},"PeriodicalIF":6.8,"publicationDate":"2026-01-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"145923680","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":2,"RegionCategory":"工程技术","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

引用次数: 0

VoxVeritasNet: A new feature engineering model leveraging iterative feature selection for detecting fake or real speech VoxVeritasNet：一个新的特征工程模型，利用迭代特征选择来检测虚假或真实的语音

IF 6.8 2区工程技术 Q1 ENGINEERING, MULTIDISCIPLINARY

alexandria engineering journal

Pub Date : 2026-01-01 Epub Date: 2026-01-14 DOI: 10.1016/j.aej.2026.01.009

Burak Çelik , Burcu Zeybek , Mahmut Burak Karadeniz , Adem Kocyigit , Onur Arsalı , Ebru Efeoglu , Bahattin Türetken

This study introduces VoxVeritasNet, a high-precision and computationally efficient feature engineering framework for deepfake audio detection. The methodology leverages a nine-level Multi-Level Discrete Wavelet Transform (MDWT) to capture intricate time–frequency artifacts. A key innovation is the quantum-inspired dual-path mapping algorithm, which models parallel signal dependencies and embeds features into a high-dimensional Hilbert space for enhancing geometric separability. To optimize performance, an iterative ensemble selection strategy utilizing Neighborhood Component Analysis (NCA), Chi

^{2}

, and ReliefF is employed alongside Support Vector Machines and k-Nearest Neighbors. The framework was evaluated across three public datasets with varying class distributions, achieving state-of-the-art peak accuracies of 99.96% with db4 and 99.71% with sym8 wavelets. Even using with the computationally efficient sym4 baseline, the model maintained exceptional detection rates above 98.99% and an equal error rate (EER) as low as 0.14%. VoxVeritasNet operates with a processing throughput of 6.45 segments per second on standard CPU hardware with a negligible storage footprint, offering a lightweight and explainable alternative to resource-intensive deep learning architectures.

本研究介绍了VoxVeritasNet，这是一个用于深度假音频检测的高精度和计算效率高的特征工程框架。该方法利用九级多电平离散小波变换（MDWT）来捕获复杂的时频伪像。一个关键的创新是量子启发的双路径映射算法，该算法模拟并行信号依赖并将特征嵌入高维希尔伯特空间以增强几何可分性。为了优化性能，利用邻域成分分析（NCA）、Chi2和ReliefF的迭代集成选择策略与支持向量机和k近邻一起使用。该框架在三个具有不同类别分布的公共数据集上进行了评估，db4和sym8小波的峰值精度分别达到了99.96%和99.71%。即使使用计算效率高的sym4基线，该模型的异常检测率也保持在98.99%以上，相等错误率（EER）低至0.14%。VoxVeritasNet在标准CPU硬件上以每秒6.45段的处理吞吐量运行，存储占用可以忽略不计，为资源密集型深度学习架构提供了轻量级和可解释的替代方案。

{"title":"VoxVeritasNet: A new feature engineering model leveraging iterative feature selection for detecting fake or real speech","authors":"Burak Çelik , Burcu Zeybek , Mahmut Burak Karadeniz , Adem Kocyigit , Onur Arsalı , Ebru Efeoglu , Bahattin Türetken","doi":"10.1016/j.aej.2026.01.009","DOIUrl":"10.1016/j.aej.2026.01.009","url":null,"abstract":"<div><div>This study introduces VoxVeritasNet, a high-precision and computationally efficient feature engineering framework for deepfake audio detection. The methodology leverages a nine-level Multi-Level Discrete Wavelet Transform (MDWT) to capture intricate time–frequency artifacts. A key innovation is the quantum-inspired dual-path mapping algorithm, which models parallel signal dependencies and embeds features into a high-dimensional Hilbert space for enhancing geometric separability. To optimize performance, an iterative ensemble selection strategy utilizing Neighborhood Component Analysis (NCA), Chi<span><math><msup><mrow></mrow><mrow><mn>2</mn></mrow></msup></math></span>, and ReliefF is employed alongside Support Vector Machines and k-Nearest Neighbors. The framework was evaluated across three public datasets with varying class distributions, achieving state-of-the-art peak accuracies of 99.96% with db4 and 99.71% with sym8 wavelets. Even using with the computationally efficient sym4 baseline, the model maintained exceptional detection rates above 98.99% and an equal error rate (EER) as low as 0.14%. VoxVeritasNet operates with a processing throughput of 6.45 segments per second on standard CPU hardware with a negligible storage footprint, offering a lightweight and explainable alternative to resource-intensive deep learning architectures.</div></div>","PeriodicalId":7484,"journal":{"name":"alexandria engineering journal","volume":"136 ","pages":"Pages 89-104"},"PeriodicalIF":6.8,"publicationDate":"2026-01-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"145975060","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":2,"RegionCategory":"工程技术","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

引用次数: 0

QuantumMedKD: A hybrid quantum–classical knowledge distillation framework for medical image analysis QuantumMedKD：用于医学图像分析的混合量子-经典知识蒸馏框架

IF 6.8 2区工程技术 Q1 ENGINEERING, MULTIDISCIPLINARY

alexandria engineering journal

Pub Date : 2026-01-01 Epub Date: 2025-12-05 DOI: 10.1016/j.aej.2025.12.007

MD Nahid Hassan Nishan , Mohammad Junayed Hasan , M.R.C. Mahdy

While hybrid quantum–classical architectures and knowledge distillation have each been explored independently in medical imaging and model compression, no prior work has unified these domains within a single framework. This study addresses this gap by investigating the integration of quantum–classical neural networks with knowledge distillation for medical image analysis, with a focus on circuit behavior, parameter efficiency, and diagnostic performance. The framework (QuantumMedKD) employs classical CNN architectures (ResNet-50, EfficientNet-B0, Xception) as teacher models, transferring learned representations to parameter-efficient quantum student networks via parameterized quantum circuits across qubit configurations (3-8), demonstrated through pneumonia detection from pediatric chest radiographs. Experimental validation reveals remarkable parameter efficiency, requiring only 24-36 trainable parameters compared to millions for classical counterparts, achieving compression ratios exceeding

1 0^{5}

while maintaining competitive diagnostic performance. Optimal configurations achieve 84.00% accuracy (EfficientNet-B0, 6-qubit) and 73.33% (Xception, 4-qubit), with knowledge distillation providing statistically significant improvements (

p

-vallues

<

0.001, Cohen’s d

>

2.0). Medical evaluation confirms diagnostic capability with sensitivity 81.4%, specificity 73.9%, and AUC 0.89, establishing quantum–classical knowledge transfer viability for resource-constrained healthcare deployment. QuantumMedKD reveals essential principles and proof-of-concept for quantum-enhanced healthcare AI systems within and beyond the noisy intermediate-scale quantum (NISQ) era for high efficiency and deployability, paving the way for future advancements in the field.

虽然混合量子经典架构和知识蒸馏各自在医学成像和模型压缩中进行了独立的探索，但之前的工作没有将这些领域统一在一个框架内。本研究通过研究量子经典神经网络与医学图像分析知识蒸馏的集成来解决这一差距，重点关注电路行为、参数效率和诊断性能。该框架（QuantumMedKD）采用经典的CNN架构（ResNet-50、EfficientNet-B0、Xception）作为教师模型，通过跨量子位配置的参数化量子电路（3-8）将学习到的表征转移到参数化的量子学生网络中，并通过儿童胸部x光片的肺炎检测进行了演示。实验验证显示了显著的参数效率，只需要24-36个可训练参数，而传统的同类方法需要数百万个，在保持有竞争力的诊断性能的同时实现超过105的压缩比。最优配置的准确率达到84.00% （EfficientNet-B0, 6量子位）和73.33% （Xception， 4量子位），知识蒸馏提供了统计学上显著的改进（p值<； 0.001, Cohen 's d > 2.0）。医学评估确认诊断能力的灵敏度为81.4%，特异性为73.9%，AUC为0.89，为资源受限的医疗部署建立了量子经典知识转移可行性。QuantumMedKD揭示了量子增强医疗保健人工智能系统在嘈杂的中等规模量子（NISQ）时代内外的基本原则和概念验证，以实现高效率和可部署性，为该领域的未来发展铺平了道路。

{"title":"QuantumMedKD: A hybrid quantum–classical knowledge distillation framework for medical image analysis","authors":"MD Nahid Hassan Nishan , Mohammad Junayed Hasan , M.R.C. Mahdy","doi":"10.1016/j.aej.2025.12.007","DOIUrl":"10.1016/j.aej.2025.12.007","url":null,"abstract":"<div><div>While hybrid quantum–classical architectures and knowledge distillation have each been explored independently in medical imaging and model compression, no prior work has unified these domains within a single framework. This study addresses this gap by investigating the integration of quantum–classical neural networks with knowledge distillation for medical image analysis, with a focus on circuit behavior, parameter efficiency, and diagnostic performance. The framework (QuantumMedKD) employs classical CNN architectures (ResNet-50, EfficientNet-B0, Xception) as teacher models, transferring learned representations to parameter-efficient quantum student networks via parameterized quantum circuits across qubit configurations (3-8), demonstrated through pneumonia detection from pediatric chest radiographs. Experimental validation reveals remarkable parameter efficiency, requiring only 24-36 trainable parameters compared to millions for classical counterparts, achieving compression ratios exceeding <span><math><mrow><mn>1</mn><msup><mrow><mn>0</mn></mrow><mrow><mn>5</mn></mrow></msup></mrow></math></span> while maintaining competitive diagnostic performance. Optimal configurations achieve 84.00% accuracy (EfficientNet-B0, 6-qubit) and 73.33% (Xception, 4-qubit), with knowledge distillation providing statistically significant improvements (<span><math><mi>p</mi></math></span>-vallues <span><math><mo><</mo></math></span> 0.001, Cohen’s d <span><math><mo>></mo></math></span> 2.0). Medical evaluation confirms diagnostic capability with sensitivity 81.4%, specificity 73.9%, and AUC 0.89, establishing quantum–classical knowledge transfer viability for resource-constrained healthcare deployment. QuantumMedKD reveals essential principles and proof-of-concept for quantum-enhanced healthcare AI systems within and beyond the noisy intermediate-scale quantum (NISQ) era for high efficiency and deployability, paving the way for future advancements in the field.</div></div>","PeriodicalId":7484,"journal":{"name":"alexandria engineering journal","volume":"134 ","pages":"Pages 49-68"},"PeriodicalIF":6.8,"publicationDate":"2026-01-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"145684093","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":2,"RegionCategory":"工程技术","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

引用次数: 0

Automatic intrusion detection and warning method in the hoisting scenario integrated BIM and GPS 集成BIM和GPS的吊装场景入侵自动检测预警方法

IF 6.8 2区工程技术 Q1 ENGINEERING, MULTIDISCIPLINARY

alexandria engineering journal

Pub Date : 2026-01-01 Epub Date: 2025-12-09 DOI: 10.1016/j.aej.2025.11.052

Xuefeng Zhao, Haodong Chen, Meng Zhang, Dechun Lu, Zhe Sun

Monitoring of intrusion at construction sites is crucial to ensure personnel safety. However, current systems struggle to automatically determine the spatial extents of hoisting areas and reliably assess worker movements in dynamic construction environments. This study proposes a novel three-dimensional (3D) automatic intrusion detection method that uniquely integrates Building Information Modeling (BIM) data with Real-Time Kinematic (RTK)-enhanced Global Positioning System (GPS) data. The proposed methodology automatically extracts BIM parameters to compute dynamic spatial boundaries of hoisting areas and converts geographic coordinates into a unified 3D virtual environment. The study’s key novelty lies in its rule-based approach that considers both worker location and movement direction to minimize false alarms, addressing a critical limitation in existing position-only detection systems. A dual alert mechanism is implemented, facilitating real-time warnings through intelligent safety helmets for field workers and a comprehensive web-based management interface for supervisors. Validation tests demonstrate substantial improvement in detection accuracy. Proposed rule-based algorithms, which incorporate both spatial position and movement direction analysis, achieved a mean error rate of 8.9 % compared to 42.8 % for traditional position-only methods tested under identical conditions. This represents a 79.2 % reduction in false alarms compared to traditional position-based methods. This scalable solution offers significant potential for enhancing personnel safety management across diverse construction sites and can be extended to monitor multiple workers simultaneously. The system’s integration capabilities make it suitable for widespread adoption in construction safety practices. However, the current implementation is limited to outdoor environments and single-worker scenarios, with future research needed to address indoor applications and multi-worker detection scenarios.

监控建筑工地的入侵对确保人员安全至关重要。然而，目前的系统难以自动确定起重区域的空间范围，也难以在动态施工环境中可靠地评估工人的运动。本研究提出了一种新颖的三维（3D）自动入侵检测方法，该方法独特地将建筑信息模型（BIM）数据与实时运动学（RTK）增强的全球定位系统（GPS）数据集成在一起。提出的方法自动提取BIM参数，计算吊装区域的动态空间边界，并将地理坐标转换为统一的三维虚拟环境。该研究的关键新颖之处在于其基于规则的方法，该方法考虑了工人的位置和移动方向，以最大限度地减少误报，解决了现有位置检测系统的一个关键限制。实施双重警报机制，通过现场工作人员的智能安全帽和主管人员的综合网络管理界面促进实时警报。验证测试表明检测精度有了实质性的提高。本文提出的基于规则的算法结合了空间位置和运动方向分析，在相同条件下测试的平均错误率为8.9 %，而传统的仅定位方法的平均错误率为42.8% %。与传统的基于位置的方法相比，这代表误报率降低了79.2% %。这种可扩展的解决方案为加强不同建筑工地的人员安全管理提供了巨大的潜力，并且可以扩展到同时监控多名工人。该系统的集成能力使其适合在建筑安全实践中广泛采用。然而，目前的实施仅限于室外环境和单工人场景，未来的研究需要解决室内应用和多工人检测场景。

{"title":"Automatic intrusion detection and warning method in the hoisting scenario integrated BIM and GPS","authors":"Xuefeng Zhao, Haodong Chen, Meng Zhang, Dechun Lu, Zhe Sun","doi":"10.1016/j.aej.2025.11.052","DOIUrl":"10.1016/j.aej.2025.11.052","url":null,"abstract":"<div><div>Monitoring of intrusion at construction sites is crucial to ensure personnel safety. However, current systems struggle to automatically determine the spatial extents of hoisting areas and reliably assess worker movements in dynamic construction environments. This study proposes a novel three-dimensional (3D) automatic intrusion detection method that uniquely integrates Building Information Modeling (BIM) data with Real-Time Kinematic (RTK)-enhanced Global Positioning System (GPS) data. The proposed methodology automatically extracts BIM parameters to compute dynamic spatial boundaries of hoisting areas and converts geographic coordinates into a unified 3D virtual environment. The study’s key novelty lies in its rule-based approach that considers both worker location and movement direction to minimize false alarms, addressing a critical limitation in existing position-only detection systems. A dual alert mechanism is implemented, facilitating real-time warnings through intelligent safety helmets for field workers and a comprehensive web-based management interface for supervisors. Validation tests demonstrate substantial improvement in detection accuracy. Proposed rule-based algorithms, which incorporate both spatial position and movement direction analysis, achieved a mean error rate of 8.9 % compared to 42.8 % for traditional position-only methods tested under identical conditions. This represents a 79.2 % reduction in false alarms compared to traditional position-based methods. This scalable solution offers significant potential for enhancing personnel safety management across diverse construction sites and can be extended to monitor multiple workers simultaneously. The system’s integration capabilities make it suitable for widespread adoption in construction safety practices. However, the current implementation is limited to outdoor environments and single-worker scenarios, with future research needed to address indoor applications and multi-worker detection scenarios.</div></div>","PeriodicalId":7484,"journal":{"name":"alexandria engineering journal","volume":"134 ","pages":"Pages 81-97"},"PeriodicalIF":6.8,"publicationDate":"2026-01-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"145735321","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":2,"RegionCategory":"工程技术","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

引用次数: 0

Accurately recognizing driver emotions through using CNN fused features and NasNet-large model 利用CNN融合特征和NasNet-large模型对驾驶员情绪进行准确识别

IF 6.8 2区工程技术 Q1 ENGINEERING, MULTIDISCIPLINARY

alexandria engineering journal

Pub Date : 2026-01-01 Epub Date: 2025-12-10 DOI: 10.1016/j.aej.2025.12.010

Khalid Zaman , Rafiullah Khan , Gan Zengkang , Sajjad Ullah Khan , Farman Ali , Tariq Hussain

This research endeavours to enhance road safety by developing an accurate driver emotion recognition system. A novel model is introduced, incorporating transfer learning principles alongside NasNet-Large CNN and Faster R-CNN, specifically designed for Driver Facial Expression (DFE) analysis. The primary objective is to bolster the recognition accuracy of Driver Facial Expression Recognition (DFER). A noteworthy improvement in the accuracy and efficiency of facial detection is attained by customizing the Faster R-CNN learning module with the Inception V3 model. The capability to accurately detect emotions is of paramount importance, as it facilitates timely interventions to avert potential accidents. To address the challenges associated with DFER accuracy in low-resolution images, this research deploys a myriad of deep learning methodologies. Through a meticulous analysis, the study identifies and implements feasible and superior solutions to enhance DFER accuracy. Additionally, the inherent constraints of low-resolution images are mitigated through the strategic application of data augmentation techniques. The evaluation of this research showcases impressive accuracy levels across diverse datasets, including JAFFE, CK+ , FER-2013, and DFERCD. These findings bear substantial implications for enhancing Advanced Driver Assistance Systems (ADAS) and contribute substantially to the overarching realm of road safety.

本研究旨在开发一套准确的驾驶情绪识别系统，以提高道路安全。引入了一种新的模型，结合了迁移学习原理以及NasNet-Large CNN和Faster R-CNN，专门为驾驶员面部表情（DFE）分析设计。主要目的是提高驾驶员面部表情识别（DFER）的识别精度。通过使用Inception V3模型定制Faster R-CNN学习模块，可以显著提高人脸检测的准确性和效率。准确检测情绪的能力至关重要，因为它有助于及时干预以避免潜在的事故。为了解决低分辨率图像中与DFER准确性相关的挑战，本研究部署了无数的深度学习方法。通过细致的分析，本研究确定并实施了可行且优越的解决方案，以提高DFER的准确性。此外，通过数据增强技术的战略应用，降低了低分辨率图像的固有限制。这项研究的评估显示，在不同的数据集，包括JAFFE， CK+ ，FER-2013和DFERCD令人印象深刻的准确性水平。这些发现对增强高级驾驶辅助系统（ADAS）具有重大意义，并对道路安全的总体领域做出重大贡献。

{"title":"Accurately recognizing driver emotions through using CNN fused features and NasNet-large model","authors":"Khalid Zaman , Rafiullah Khan , Gan Zengkang , Sajjad Ullah Khan , Farman Ali , Tariq Hussain","doi":"10.1016/j.aej.2025.12.010","DOIUrl":"10.1016/j.aej.2025.12.010","url":null,"abstract":"<div><div>This research endeavours to enhance road safety by developing an accurate driver emotion recognition system. A novel model is introduced, incorporating transfer learning principles alongside NasNet-Large CNN and Faster R-CNN, specifically designed for Driver Facial Expression (DFE) analysis. The primary objective is to bolster the recognition accuracy of Driver Facial Expression Recognition (DFER). A noteworthy improvement in the accuracy and efficiency of facial detection is attained by customizing the Faster R-CNN learning module with the Inception V3 model. The capability to accurately detect emotions is of paramount importance, as it facilitates timely interventions to avert potential accidents. To address the challenges associated with DFER accuracy in low-resolution images, this research deploys a myriad of deep learning methodologies. Through a meticulous analysis, the study identifies and implements feasible and superior solutions to enhance DFER accuracy. Additionally, the inherent constraints of low-resolution images are mitigated through the strategic application of data augmentation techniques. The evaluation of this research showcases impressive accuracy levels across diverse datasets, including JAFFE, CK+ , FER-2013, and DFERCD. These findings bear substantial implications for enhancing Advanced Driver Assistance Systems (ADAS) and contribute substantially to the overarching realm of road safety.</div></div>","PeriodicalId":7484,"journal":{"name":"alexandria engineering journal","volume":"134 ","pages":"Pages 177-196"},"PeriodicalIF":6.8,"publicationDate":"2026-01-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"145735352","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":2,"RegionCategory":"工程技术","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

引用次数: 0

SR-PAN-EffNet: A collaborative optimization approach for low-quality image classification and restoration SR-PAN-EffNet：一种低质量图像分类与恢复的协同优化方法

IF 6.8 2区工程技术 Q1 ENGINEERING, MULTIDISCIPLINARY

alexandria engineering journal

Pub Date : 2026-01-01 Epub Date: 2025-12-24 DOI: 10.1016/j.aej.2025.12.029

Jun Li , Dong Liu

Low-quality images, due to degradation issues such as noise and compression artifacts, often lead to feature extraction distortion in classification models, thereby reducing classification accuracy. This issue is particularly prominent in practical computer vision applications. To address this, this paper proposes the SR-PAN-EffNet model, which integrates a semantic-guided SRGAN restoration module with a noise-aware PANet attention module, and employs end-to-end joint training to achieve the collaborative optimization of “restoration serving classification”. Experimental results show that the model achieves Top-1 accuracy of 68.5% and 57.6% on the ImageNet-1K and CIFAR-100 low-quality datasets, respectively, improving by 4.7–6.4 percentage points over NFNet-F4, with PSNR and SSIM also leading. Ablation experiments reveal that removing core modules results in a 2.9–6.8 percentage point decrease in accuracy. Future work will focus on improving the model’s real-time performance and robustness through lightweight design, extreme degradation adaptive optimization, and category-adaptive guidance, aiming to promote its application in scenarios such as surveillance and medical imaging.

低质量图像由于噪声和压缩伪影等退化问题，往往会导致分类模型的特征提取失真，从而降低分类精度。这个问题在实际的计算机视觉应用中尤为突出。针对这一问题，本文提出SR-PAN-EffNet模型，该模型将语义引导的SRGAN恢复模块与噪声感知的PANet关注模块集成在一起，采用端到端联合训练实现“恢复服务分类”的协同优化。实验结果表明，该模型在ImageNet-1K和CIFAR-100低质量数据集上分别达到了68.5%和57.6%的Top-1准确率，比NFNet-F4提高了4.7-6.4个百分点，其中PSNR和SSIM也处于领先地位。烧蚀实验表明，去除核心模块会导致精度降低2.9-6.8个百分点。未来的工作将重点通过轻量化设计、极端退化自适应优化和类别自适应制导来提高模型的实时性和鲁棒性，以促进其在监控和医学成像等场景中的应用。

{"title":"SR-PAN-EffNet: A collaborative optimization approach for low-quality image classification and restoration","authors":"Jun Li , Dong Liu","doi":"10.1016/j.aej.2025.12.029","DOIUrl":"10.1016/j.aej.2025.12.029","url":null,"abstract":"<div><div>Low-quality images, due to degradation issues such as noise and compression artifacts, often lead to feature extraction distortion in classification models, thereby reducing classification accuracy. This issue is particularly prominent in practical computer vision applications. To address this, this paper proposes the SR-PAN-EffNet model, which integrates a semantic-guided SRGAN restoration module with a noise-aware PANet attention module, and employs end-to-end joint training to achieve the collaborative optimization of “restoration serving classification”. Experimental results show that the model achieves Top-1 accuracy of 68.5% and 57.6% on the ImageNet-1K and CIFAR-100 low-quality datasets, respectively, improving by 4.7–6.4 percentage points over NFNet-F4, with PSNR and SSIM also leading. Ablation experiments reveal that removing core modules results in a 2.9–6.8 percentage point decrease in accuracy. Future work will focus on improving the model’s real-time performance and robustness through lightweight design, extreme degradation adaptive optimization, and category-adaptive guidance, aiming to promote its application in scenarios such as surveillance and medical imaging.</div></div>","PeriodicalId":7484,"journal":{"name":"alexandria engineering journal","volume":"134 ","pages":"Pages 542-555"},"PeriodicalIF":6.8,"publicationDate":"2026-01-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"145837445","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":2,"RegionCategory":"工程技术","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

引用次数: 0