Panoptes

2021-2022

Panoptes is a web-based application which detects anomalies from live camera footage and then reports these anomalies to Authorities. These anomalies can be Road Accident, Vandalism, burglary, etc.

Introduction

With technology advancing, we want technology that makes life easier for us. We want systems that will act as humans so that we can focus on something more productive. We should have systems that will automate things that can be tedious for us. Watching live feed is something tedious but most of the places still have no automated systems for it. This is what we want to develop, an automated system that will make detect anything suspicious from the live stream of security surveillance cameras. The system can distinguish normal activities from suspicious activities and report them to the correct authorities.

The topic of this study will automate tedious work as monitoring live streams at malls. Malls and different stores hire security personnel whose job is to just watch a live stream from security cameras. This job can become tedious which can lead to negligence because 95% of the time, there is nothing to look at, which can lead to the person neglecting the live stream. Now, this negligence might cost as the person might neglect the stream at a moment when there is some suspicious activity occurring. On the other hand, a system can work 24/7. This can lead to automated systems which will reduce costs for organizations that they could utilize somewhere productive.

Project Scope

The general idea of our project is that this project will help the organizations like Punjab Safe City Authority (PSCA) which is a Pakistani autonomous government body whose purpose is to make Punjab and soon the whole Pakistan safer from criminal activity. Organizations like these have access to all the cameras installed in a specific region which is whole Punjab in the case of PSCA. All these cameras are constantly monitoring the streets of Punjab and in all fairness, it is quite difficult, costly, and impractical to hire personnel to watch the live stream from each camera. Which means these cameras are pretty much useless for watching criminal activity at real time. Although these cameras can be quite helpful to watch the videos of criminal activities later for investigation or any other reason, but they are not quite helpful for detecting crime at real time.

Our project aims to make these cameras useful at detecting crime in real time. Our system will constantly analyze the live streams and will detect anomalous activities. If a camera detects an anomalous activity, it reports this anomaly to the department which should deal with the anomaly like in case of arson, our software will notify the fire department and in case of road accident, our software should notify Hospital and Police department.

Objectives

The objectives for our proposed project will be as follows:

● The software will constantly analyze video streams from security cameras and will look for anomalous activities.

● The software will recognize activities that are anomalous and which activities are not.

● If there is an anomalous activity detected, then it will be reported to proper authorities.

● The software will have a location of each security camera and will send the location of the camera on which the video of anomaly activity was captured.

● The software will also cut a segment of a live stream in which suspicious activity is occurring and then send the video to the authorities as well.

● Both location and video segments will be sent together along with the type of anomaly (i.e. abuse or violence or road accident, etc.)

● The software will also save these clips in a repository so that they can be viewed later.

Working

Cameras which will be connected to the server will send the live stream to the machine learning algorithm which will detect the anomaly. After Saving the anomaly in the database, server will send a notification to the Authorities which will arrive at the scene of anomaly