Accident Detection and Alert System

Abstract - One of the leading causes of death and serious injury all over the world is due to road traffic accidents, especially in regions with high population and mobility like urban areas. There are lots of cities with surveillance systems like CCTV, but the instantaneous detection of an accident so that timely emergency response is possible is still a problem due to manual watching or analysis of footage after the fact. This research proposes an AI-based Accident Detection and Emergency Notification System, a smart, self sufficient platform that analyzes deep learning techniques to detect accidents on the highways and autonomously notify the appropriate personnel through SMS, email, and honking programmable speakers at the scene. Unlike other systems that still rely on humans to monitor or sensors within vehicles to alert, our system utilizes the traffic surveillance camera to real-time feed the video CCTV frame to a CNN model which classifies each frame as Accident or No Accident. The system's multi-channel alerting system sends SMS alerts to emergency responders and traffic control centers using Twilio, automated emails with SMTP, and plays an audible alarm with playsound library for Open CV. This guarantees rapid and extensive communication to enable prompt operational response coordination. Key Words: SMS, SMTP, Twilio, OpenCV, Computer Vision, Flask, Real-time Monitoring, CNN, Surveillance Cameras, Emergency Alert Systems, AI-based Accident Detection, Deep Learning