With the current development of technology, CCTV surveillance is widely used as a method of observing uncivilized public behavior, traffic-related violations, and manmade disasters. With the increase of human misbehavior, traffic on roads, and fire-related accidents in Sri Lanka, automated video tracking of those scenarios is the immediate need. Automation is crucial as the random nature of the above-stated incidents. As the goal of this research, a novel approach, a CCTV-based public security system that is capable of automatically detecting Fires, Road Accidents, Traffic rule violations, and Civil Unrest is proposed to implement using Image processing and Machine Learning. Having an automated security system like this will pave the way to reduce the errors that occur due to manual approaches. This document provides an outline of the proposed system and a summary of outcomes authors have achieved when training appropriate machine learning models to work with real-time data. Systems’ functionality includes identifying initial objects through object detection like humans, vehicles, white lines and correctly detect and localizing accidents, white line violation on roads, the occurrence of fires, and fighting related to civil unrest from incoming video frames which belong to CCTV live steam. This system also generates real-time alerts for main controlling admins when the implemented model detects such scenarios.

Machine Learning Automated Security System Real-Time Alerts.