Abstract: Accurate, continuous vehicle traffic data is fundamental to evidence-based road pavement design, yet conventional manual counting methods remain incapable of meeting this need in high-density traffic corridors such as the Nsukka–Enugu Road in southeastern Nigeria. This paper presents the design, implementation, and evaluation of an end-to-end Vehicle and Number Plate Recognition (VNPR) system that integrates YOLOv8 Convolutional Neural Network detection, EasyOCR optical character recognition (OCR), OpenCV-based image preprocessing, and Supabase PostgreSQL cloud database management to automate real-time vehicle identification and traffic logging. The system was trained on a combined dataset of 12,746 annotated images (Kaggle open repository and field-collected images from the Nsukka–Enugu Road corridor) and validated using Levenshtein distance-based character accuracy metrics. The complete software pipeline — from frame acquisition and preprocessing through bounding-box detection, plate cropping, OCR extraction, and database insertion — is described in architectural detail. OCR evaluation on standard test images yielded perfect exact-match accuracy (1.0) in optimal conditions, with systematic character confusion errors (0/O, 1/I, 5/S) identified as the dominant failure mode under live traffic conditions. Traffic records collected through the system were applied to compute Average Daily Traffic (ADT) and Equivalent Single Axle Loads (ESALs) per the Nigerian Highway Manual, yielding a 20-year cumulative ESAL of 10,974 and informing a three-layer pavement specification comprising lime-stabilised subgrade, 200 mm granular base, and 100 mm dense-graded asphalt. A companion mobile application supporting real-time tracking, cross-checking, and database interaction completes the system. The architecture and findings presented are directly transferable to road infrastructure planning in Nigeria and similarly resource-constrained traffic environments.
Keywords: YOLOv8; EasyOCR; OpenCV; vehicle number plate recognition; deep learning; Supabase PostgreSQL; road pavement design; ESAL; Nsukka–Enugu Road; Nigeria.
Title: An Integrated System for Vehicle and Number Plate Identification Using Deep Learning and Database Management for Enhanced Road Pavement Design: Architecture, Implementation, and Evaluation on the Nsukka–Enugu Road, Nigeria
Author: Oyesanya Oluwatosin Gabriel
International Journal of Civil and Structural Engineering Research
ISSN 2348-7607 (Online)
Vol. 14, Issue 1, April 2026 - September 2026
Page No: 46-56
Research Publish Journals
Website: www.researchpublish.com
Published Date: 15-June-2026
