INTELLIGENT VIDEO SURVEILLANCE USING DEEP LEARNING SYSTEM

Author: Dr Prabanand S C, Sridevi T, Dinesh L, Subrajashree D, Sridhar C

Published On: 2024-12-02

Intelligent video surveillance systems have become vital for enhancing road safety by providing real-time monitoring and rapid response to critical incidents. This paper presents a deep learning-based system specifically designed for accident detection in high-traffic areas. Utilizing advanced models such as YOLOv7 for real-time object detection and RCNN for accurate bounding box generation, the system efficiently identifies vehicles and collision events from video feeds. FFmpeg is used for seamless video processing, enabling continuous monitoring in diverse conditions. To analyze the temporal sequence of events leading up to an accident, LSTM networks are employed, providing robust video analysis and context understanding. Once an accident is detected, the system leverages Twilio to send immediate notifications to nearby police and ambulance services, significantly reducing response time. This intelligent surveillance solution addresses critical challenges such as real-time detection, false-positive reduction, and environmental variations, enhancing both the accuracy and effectiveness of accident detection. The study also outlines the limitations and future directions for improving intelligent video surveillance in public safety applications. Index terms - Accident detection, deep learning, YOLOv7, RCNN, FFmpeg, LSTM, Twilio, Video surveillance, Emergency response, real-time monitoring