Smart surveillance advance system using machine learning and CNN to learn suspicious activity

Abstract

Nowadays we can find security cameras everywhere, from roads to malls, we are under surveillance. A lot of people install security cameras in their houses and the sole reason for installing security cameras is safety. Everyone is worried about the safety of someone, some people are worried about themselves, some about their family, and some are concerned about the safety of their friends. In short, everyone cares about being safe. But how about increasing the security up a notch? How about people installing software with their security cameras that will automatically detect suspicious activities and report authorities? The primary goal of this Research is to develop a system that is capable to detect any anomalous and suspicious activities from given video frames captured by surveillance cameras and reporting them to authorized personnel for immediate actions. It is very common in surveillance that anomalies go unnoticed by the guards on the spot. Thousands of cameras are installed on streets and roads that record everything, but that recording is only used later because no one was watching a live video stream at that exact moment. But what if the software is always watching these thousands of live video streams? This means that every anomalous activity or crime that occurs and is recorded by these cameras will be reported instantly. This software will use an advanced Machine learning model to recognize suspicious activities and report them to proper authorities.