Loading...
Edge AI-Powered Real-Time Speed Estimation and License Plate Recognition for Urban Road Safety
Author(s)
Asavanarakul, Prach
Asavanarakul, Aran
Ampha, Phongsuk
Ounthong, Wissaroot
Namracha, Dutsadi
Khongsuwan, Noppharat
Date Issued
January 1, 2025
Type
Conference Paper
Abstract
This paper presents a real-time vehicle speed estimation and license plate recognition (LPR) system designed for urban and community deployment. The proposed system integrates a Thai-optimized dual-AI LPR framework with high-precision FMCW millimeter-wave radar, replacing earlier microwave radar and traditional image processing. A 30 km/h threshold is applied, with radar-triggered detection within 10-20 meters operational range, automatically capturing images when vehicles exceed limits. License plate detection utilizes YOLOv8s, followed by YOLOv8m for character recognition with coordinate-based sequencing, achieving markedly superior performance compared with OpenCV-based approaches. The architecture is implemented on the NVIDIA Jetson Xavier platform, enabling sub-100ms edge processing without a centralized server dependency. Experimental evaluation on 2000 vehicle samples demonstrates 94.95% accuracy for combined speed estimation and Thai license plate recognition within a 30-60 km/h range. Real-time outputs, including timestamps, plate numbers, and speeds, are logged and displayed on roadside monitors for driver feedback. Thailand faces critical traffic enforcement gaps, with 72.17% of accidents occurring in local areas and 25.4 fatalities per 100,000 population. The proposed system addresses this challenge through cost-effective edge-AI deployment for community traffic management. Limitations include untested performance under low-light, nighttime, and adverse weather conditions. These areas will be investigated in future work.
Citation
Temscon Aspac 2025 4th IEEE Technology and Engineering Management Conference Asia Pacific, 2025