Automated optimal parking slot prediction using deep learning and digital twin technology aided parking space management for material science application

As vehicles on the roadside increase exponentially, drivers find it complicated to recognize parking areas. This makes it essential to identify an optimized model for resolving the vehicle-parking issues. In other words, a practical model must be implemented to identify outdoor parking slot status u...

وصف كامل

التفاصيل البيبلوغرافية
الحاوية / القاعدة:ALEXANDRIA ENGINEERING JOURNAL
المؤلفون الرئيسيون: Lu, Ke; Zheng, Bei; Shi, Jingjing; Xu, Yaowen
التنسيق: مقال
اللغة:English
منشور في: ELSEVIER 2025
الموضوعات:
الوصول للمادة أونلاين:https://www-webofscience-com.uitm.idm.oclc.org/wos/woscc/full-record/WOS:001447049000001
الوصف
الملخص:As vehicles on the roadside increase exponentially, drivers find it complicated to recognize parking areas. This makes it essential to identify an optimized model for resolving the vehicle-parking issues. In other words, a practical model must be implemented to identify outdoor parking slot status using sensing material or vehicles. For this purpose, the proposed technique aims at presenting an automated optimal parking slot detection and management using Active Learning (AL) and deep learning-based prediction model. The input images are retrieved from the input image dataset (PkLot). Then, the preprocessing stage is carried out by resizing, image enhancement, background subtraction, Hough transform, and a mixture of Gaussians. Improved Pre-trained UNet-based feature extraction is carried out. The optimal features are selected using the Modified chaotic BAT optimization approach. The classification is finally done using Deep Cascaded Fine-tuned Active Learning and Inception V3 technique. The results are contrasted with the suggested approach and existing methods. The detected result is stored in the server regarding the real-time availability of slots. Then, digital twin technology manages parking space management to ensure slot availability. The assessment of performance is evaluated for varied metrics like mean accuracy, sensitivity, specificity, F1-score, recall, precision, FNR, and FPR and outcomes are compared with existing methodologies to validate the efficacy of proposed model.
تدمد:1110-0168
2090-2670
DOI:10.1016/j.aej.2025.03.019