ANIAKOR, MAXWELL KOSISOCHUKWU (2025) ACCIDENT RISK DETECTION MODEL FOR ENUGU ROADS. Other thesis, GODFREY OKOYE UNIVERSITY.
![[img]](https://eprints.gouni.edu.ng/style/images/fileicons/text.png) |
Text
REPO240.docx Download (1MB) |
Abstract
Road traffic accidents remain a critical public safety concern across many African countries, resulting in extensive loss of life, economic strain and social disruption. Traditional accident prevention strategies, which rely heavily on post-incident analysis, have proven insufficient in addressing the rising frequency and severity of traffic-related incidents. This study proposes a real-time accident risk detection system leveraging the YOLOv8 deep learning model trained on a custom dataset tailored to African road conditions, including underrepresented vehicles like tricycles (Keke). The dataset, derived from real-world surveillance footage, was carefully preprocessed and annotated to capture diverse vehicle types, behaviors and environmental factors. YOLOv8 was employed for object detection and behavioral analysis, while additional modules analyzed temporal patterns to predict accident-prone scenarios. Evaluation metrics demonstrated high object detection accuracy ([email protected] of 86.7%) and effective accident prediction performance (overall accuracy of 84.6% and AUC of 0.89). Real-time performance testing showed a stable frame rate of 24–30 FPS with minimal latency. The system exhibited robust functionality across varying lighting, weather and traffic conditions, confirming its viability for deployment in African urban and semi-urban environments. This research highlights the transformative potential of AI-driven approaches for proactive road safety management in low-resource settings.
| Item Type: | Thesis (Other) |
|---|---|
| Subjects: | C Auxiliary Sciences of History > CE Technical chronology. Calendar |
| Divisions: | Faculty of Engineering, Science and Mathematics > School of Electronics and Computer Science |
| Depositing User: | Uchenna Eneogwe |
| Date Deposited: | 29 May 2026 13:32 |
| Last Modified: | 29 May 2026 15:31 |
| URI: | http://eprints.gouni.edu.ng/id/eprint/5679 |
Actions (login required)
 |
View Item |
