Accident Detection and Severity Classification System using YOLO Model
K. Jaspin, Allison Bright, M L Legin
Abstract
Accidents on the road can happen at any time and may result in varying degrees of severity. It is essential to have an automated system that can swiftly detect and classify these accidents and trigger appropriate responses. This study proposes an automated system for detecting and classifying road accidents, utilizing the YOLOv5 algorithm for accident detection and machine learning for assessing accident severity. YOLOv5’s efficiency for accident detection stems from its optimized architecture, improved training techniques, better generalization capabilities, and simplified deployment process. These factors collectively contribute to faster and more accurate detection of accidents, enabling timely interventions and improving road safety. This system can identify three levels of accidents: mild, moderate, and severe, triggering different responses accordingly. In the event of a mild accident, the system contacts Rescue System, a centralized monitoring center or dedicated authority for traffic and safety, which then notifies the nearby police station. For moderate accidents, the system notifies multiple authorities, including the ambulance service, fire station, and police station, ensuring a swift and comprehensive response. Severe accidents prompt a multi-pronged response, involving immediate medical assistance, potential rescue operations, and accident documentation and traffic management. This proposed system, powered by the YOLO algorithm and machine learning, provides an efficient approach to respond to road accidents by categorizing them and initiating appropriate actions. The results demonstrate an impressive average precision (AP) of 93.02% for car accident detection. Comparative analysis with alternative object detection models underscores the superior accuracy and real-time capabilities of this YOLO model. This comprehensive performance enhancement has the potential to save lives and minimize the impact of road accidents.