Automated Illegal Parking Fine System Using YOLOv8 and PaddleOCR on the Cloud

View at publisher Summarise with VIRAT
Authors: Aryan Paode, Krishna Samdani, Tanisha Verekar
Cited by: 0
References: 12

Summary

No summary yet. VIRAT can read this paper and write a plain-language summary of its aim, methods, key findings and limitations.

Summarise with VIRAT Free to try with your Science Catalogs account.

Abstract

Illegal parking in highly populated metropolitan city environments is a persistent urban challenge that leads to traffic congestion, higher CO2 emissions, and road safety risks. Existing enforcement mechanisms largely depend on manual monitoring or static image-based automated detection systems that are often inconsistent and unable to distinguish between temporarily halted vehicles and genuine parking violations. In order to address these limitations, this paper presents a cloud-enabled, AI-driven model for automated illegal parking detection that incorporates temporal validation into the decision process. The proposed system integrates YOLOv8-based vehicle detection with PaddleOCR-based license plate recognition and introduces a dualtimestamp, multi-video comparison strategy to evaluate vehicle existence at a fixed place over a time interval. This temporal validation mechanism, with fair reliability, discriminates between illegally parked vehicles and short-duration stops caused by traffic signals, congestion, or slow-moving traffic. This significantly reduces false positives common in snapshot-based approaches. The system is deployed on Amazon AWS EC2 for computation and S3 bucket for centralised storage and archival. Real-world video data collected under varied lighting conditions were used for experimental tests. The proposed model achieves an overall detection accuracy of 90.77%, demonstrating a fairly acceptable performance across daytime, low-light, and glare-prone conditions. Test results and error analysis confirm that incorporating temporal validation improves detection reliability without imposing high computational overhead. The model is designed with ethical considerations and consent-based data collection for each video. The proposed model, since developed and tested on real-world data, offers a practical, scalable, and deployable solution for smart-city traffic enforcement, with the potential to reduce manual workload.

Fields of Study & Subjects

Vehicle License Plate Recognition Smart Parking Systems Research Advanced Neural Network Applications

Publication details

Journal / source
International Conference on Smart City and Information System (ICSCIS)
Pages
173–177
Publication date
30 January 2026
Language
EN