Rapid Detection Of Crime Using Deep Learning On Fog Nodes

Project Title

Rapid Detection Of Crime Using Deep Learning On Fog Nodes

Project Area of Specialization Artificial IntelligenceProject Summary

In recent years, surveillance systems have become one of the most applied systems throughout the world in order to prevent crimes in public places, there is an increasing
need of proof based approach that could help in catching the culprit easily. Cases of crimes and harassment are increasing day by day. Both researchers and commercial community are taking a great interest in it due to its potential to reduce the crime rate. We see it deployed in big firms, super markets and top organizations all over the world. It comes to the aid of law enforcing agencies like police, to investigate crime. But right now the system works in a rather inefficient way i.e. the crime is reported by people themselves and police responds to it after the incident has occurred, by obtaining video feeds through the surveillance system and performing analysis on it or even if there is surveillance system installed there requires a need of security personal to keep a constant eye on the system which is not efficient. In a situation of life and death we need an efficient system that can generate alert almost the same time as a crime is detected. We aim to make a surveillance system that can auto detect a crime and
captures the identity of people causing it, at real time. Our system will have the records of criminals stored in its database so that even that criminal is not involved in any crime at that moment but caught on the system, it will generate the alert. Since we are working in the context of big data the data generated by our system will be of huge amount and all the processing needs to be done in real time for the system to be efficient and for real time performance, we intend to use a distributed architecture using fog nodes. The use of fog nodes is to divide all the data coming through video stream between the nodes so that there is reduced latency and faster processing. We also used cloud storage in order to backup our data continuously to perform incremental training.

Project Objectives

The objectives of this project are described below.
The number of crimes occurring on daily basis is increasing day by day and the existing encountering systems are general cctv footages which are not so efficient
given that the footages are obtained later after the incident has occurred. And because there is huge amount of video feeds it is inefficient to look for crime in those feeds. The main objective of our system is to introduce a crime detection system that could perform at real time. Secondly, the system must be able to detect the criminals whose records are
already stored in system database even when they are not found in any such activity.

Project Implementation Method

To solve the problem of crime detection at real time, we will implement the most state of the art object detector algorithm and facial recognition algorithm that give the best result as early as possible. Since the data would be distributed between multiple nodes, we will deploy a secured tunnel transfer network on infini band that transfers the data instantly to perform at real time.
The videos will be first sent to object detector where the object detection will be performed and guns and persons will be detected. The captured frames will be distributed on a certain threshold that would not cause delay for the frames to be stacked up and sent to multiple nodes using infini band network. All the nodes will have facial recognizer already in running state and perform face recognition on those received frames. If a person is identified which exists in the criminal database then the system will generates an alert.

Benefits of the Project

Following are the benefits of this project:

• The successful implementation of this project could lead to reduction of crime rate in our country.
• This system can be deployed in shopping mall, streets, buildings and in many other public places.
• This system will be able to detect the criminal even when he/she is not found with any harmful artifact like gun provided that he/she already exists in the criminal database of the system.
• And finally, this system could prove to be a lot of help for the armed forces like police department, security department of our country.

Technical Details of Final Deliverable

The final deliverable of our project would be a Research Paper/Report that will contain the
system architecture and the results of experiments performed on our rapid crime detection
system.

Final Deliverable of the Project HW/SW integrated systemType of Industry IT Technologies Artificial Intelligence(AI), Big DataSustainable Development Goals Industry, Innovation and InfrastructureRequired Resources

Item Name	Type	No. of Units	Per Unit Cost (in Rs)	Total (in Rs)
			Total in (Rs)	70000
Graphic card	Equipment	1	70000	70000