Enhancing Car Accident Detection: A Fuzzy Logic-Based Evaluation of Faster R-CNN and YOLOV8

View at publisher Summarise with VIRAT
Authors: Shaik Basheer, Chanchal Biswas, Dunga Priyatham, Shaik Abdul Karim, Abdul Rehan
Cited by: 0
References: 17

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

Real-time accident detection is essential for enhancing road safety and enabling quick emergency response. Models such as Faster R-CNN and YOLOv8 already offer strong performance, achieving detection accuracies of 80.1 % and 99.5 % (mAP@0.5), respectively. However, in challenging real-world scenarios-such as low visibility, occlusions, and overlapping objects-these models may produce inconsistent or conflicting predictions.To overcome this limitation, we propose a hybrid accident detection system that combines YOLOv8 and Faster R-CNN using a fuzzy logic-based decision mechanism. Confidence scores from both models are processed through fuzzy membership functions-Ignore, Uncertain, and Strong Detection-and a rulebased inference engine to produce a more stable and interpretable final output.Experiments on a custom dataset containing Car, Car Accident, and Fire Accident classes show that the hybrid approach achieves an improved mAP @ 0.5 of 90.3 %, reduces false positives, and delivers more reliable classification. These results demonstrate that fuzzy logic fusion effectively enhances the consistency and robustness of accident detection, making the system suitable for intelligent surveillance and real-time traffic monitoring.

Fields of Study & Subjects

Advanced Neural Network Applications IoT and GPS-based Vehicle Safety Systems Autonomous Vehicle Technology and Safety

Publication details

Journal / source
International Conference on Communication, Computing and Electronics Systems (ICCCES)
Pages
515–522
Publication date
21 January 2026
Language
EN