A Real-Time Road Accident Detection and Alert System using YOLO and Machine Learning
Road traffic accidents represent a significant global issue, the rates of which are significantly raised by the delay in emergency response existing approach are not automated and do not have advanced real-time analysis because they rely either on sensor-based detection of impact or in the CCTV monitoring. A machine learning-based road accident detection and alert system based on the YOLOv5 is proposed in the present research. Under one system, it is a mix of motion tracking and vehicle detection, estimating an accident and classifying it. Whereas MOSSE tracking is applied to maintain vehicle identity across frames, YOLOv5 is employed in vehicle detection. Crashes are estimated using velocity variation analysis and intersection over union. The Support Vector Machine is applied to identify and classify the violent flow motion descriptors to confirm collision events. Upon confirmation, crash film is stored to be monitored and analyzed, and automated email and text messages are sent to provide alerts. The technology is suitable in intelligent traffic surveillance application because experimental testing demonstrates a stable detection performance with a rapid response time.