Suspicious Person Detector

Project Title

Suspicious Person Detector

Project Area of Specialization Artificial IntelligenceProject Summary

Video surveillance Systems use CCTV cameras to transmit video signals to a specific place on a limited set of monitors. Since the need of security, Video surveillance Systems have become increasingly fast and smart. However, CCTVs are still constrained to monitoring where they can only capture and store things. They are unable to detect objects where they can notify us at the time of capturing.

For Example, Security cameras inside an Apple store in California captured a group of brazen thieves in action. In the matter of seconds, they stole 26 items, worth total of $27,000. Police has videos of robbery but the thieves didn’t get caught.

We want to make an intelligence Video surveillance System to detect a suspicious person and to asses if there’s real danger under a given scenario and generate an alert. We then send alert signals wirelessly to outside security organization i.e. police making them aware of any crime.

Such crimes are committed using gun power where criminals freeze the environment. Therefore, we consider a person to be suspicious under the following three algorithms

1. Handgun detection
2. Masked face recognition
3. Abnormal Activity recognition

Our project can be deployed in areas like banks. Security cameras installed in different locations will give their output to the detection system which in terms will constantly monitor the environment. Any person with the above three clues be detected as suspicious.

Project Objectives

1. Basic Objective of this project is the system should generate an alert whenever a person is detected as suspicious.
2. There must be proper dataset to train our network model.
3. As the detection is concerned, there must be less errors in order to minimize the rate of false negatives.
4. The system should be fast enough to work in Real time.
5. The system should differentiate between a suspicious person and normal person i.e. security person holding a gun shouldn’t be detected. Employee with a mask due to some reason shouldn’t be detected as suspicious.

Project Implementation Method

In this Project, we will use a neural network model using machine learning to detect objects in an image. We will use a real time object detector with high accuracy i.e. YOLO, fast R-CNN or faster R-CNN etc. We will then train our model on custom data for detection i.e. handguns, masks etc.

Working

Our Project contains a fixed camera to capture environment. We then extract frames of the video. We then pass frames into our network model as input. The model first will detect human if there’s any. After successful detection, it tries to detect the following three things

1.  Whether that person has weapon? (Handgun Detection)
2. Whether that person is wearing a mask?
3. Whether that person does an activity that differs from everyday activity?

After successful processing, these three detectors will give us float values from 0 to 1 which in terms will be combined together in a certain way that will generate one final value. We then compare that value to some threshold value which decides if the person is suspicious.

Benefits of the Project

Project helps in detecting suspicious person in certain areas like bank, gate passes etc. which enhances security

purposes. Unlike manual activation Project automatically activate security system whenever suspicious activity is

caught by the camera. We have learnt that Human life is more important than a machine life. Keeping this in context, we implement such system that can be beneficial for sensitive areas like banks where chances for crimes are very high. Criminals freeze the environment with the power of their gun and they do whatever they want to do. In such situations, at least we’ll have an automated alarm system.

Technical Details of Final Deliverable

Final Project Deliverable will contain Documentation and Hardware System.

Documentation includes

1. Data sets used for training the model
2. Research work
3. Implementation methods and code

Hardware includes Surveillance system with a camera working with back-end hardware.

Final Deliverable of the Project HW/SW integrated systemType of Industry IT Technologies Artificial Intelligence(AI)Sustainable Development Goals Decent Work and Economic GrowthRequired Resources

Item Name	Type	No. of Units	Per Unit Cost (in Rs)	Total (in Rs)
			Total in (Rs)	48000
Portable Smart Camera	Equipment	1	10000	10000
Wireless Network Adaptor for Raspberry Pi	Equipment	1	10000	10000
High End GPU Card for PC	Equipment	1	20000	20000
Raspberry Pi 3	Equipment	1	8000	8000