Current  Method of Surveillance is manual monitoring of CCTV footages for security purposes and manually raising alarm in case of any mishap. In case of any crime committed, dozens of security footages need to be scanned to get information about perpetrator. To further track the perpetrator more videos need to be scanned this is a labour and time intensive task. As currently it is totally manual task to analyse and monitor video footages this leads to errors and mistakes which could lead to mishaps and accidents .

What it does

Our Objective is to create a multi-camera smart surveillance network using computer vision which would be a completely software based solution with no hardware involvement.

This surveillance network would :

1.) Detect and Identify people in CCTV footage and save their data(i.e name, time, camera in which detected) in our database 2.) Identify prohibited zones and raise alarm when someone enters it 3.) The data gathered from identifying person could be used for various purpose like Tracking, Searching, etc. 4.) Another unique feature that we are introducing in CONTACT TRACING in unprecedented time's like now our network would detect people when come in contact with each other and save it in our data base and Trace all the people who came in contact with the specified person. 5.) Provides for a good visualization using heatmap to control and monitor large gathering of people within an area.

We are planning to create a complete fully functional surveillance system that would be future of surveillance which would remove all the manual efforts of going through video footages if any crime takes place and track the person through multiple camera's manually .This is a very tiresome job. Our surveillance network will significantly reduce the efforts as after the CCTV footage is processed we will have all the data of which people where present at the time of crime was committed. And if person who committed crime is found you could just search in database the person's name and you would easily get all the data of where the person went at what time and came in contact with which people. Thus reducing a lot of manual labour.

How we built it

We built it using python using YOLOV3 for detection of people and Deep Sort for tracking them in footage while using Facial recognition if face is visible and if face is not visible using Reidentification to identify the person and then storing the data in the database. The COVID 19 contact tracing would also be done and result would be presented in textual part as well as graphical form using Igraphs.

Accomplishments that I'm proud of

This project is one of a kind project that has not been build before with extra feature of contact/criminal tracking that reduces the manual labour and time taken in current surveillance methods.

Built With

Share this project: