FPGA Based Real Time Face Authorization System

Project Title

FPGA Based Real Time Face Authorization System

Project Area of Specialization Artificial IntelligenceProject Summary

The idea of this project is to modify the existing face recognition and implementing it into smart real time applications. The commonly used face recognition system are mostly PC based which are bulky, expensive and unportable. In our project we are converting PC based systems in to an embedded system. The processor used in this project can process most complex face recognition algorithm with the large database in embedded system in order to get the optimal response with least delay and can be implemented in real time environment in various application related to authorization and verification. Face recognition system is an already existing field of biometric security. Facial algorithms has been implemented in various areas of biometric utilities like  Facebook and iPhone. However, they are far from achieving the ideal in all applications utilizing these techniques in practical life. Our project isn’t just for apps(software) and social media it can also be used with database, in addition to that, it works with real time videos and is not limited to software oriented environment, Unlike Facebook and IPhone. Moreover, as it’s mentioned above that before this other techniques were used, but couldn’t get off from nail of “delay factor” .Our goal is that to bring this face recognition technology in to real life implementation. The project is to create a device that can perform a face recognition task with accuracy, least delay and larger databases so the real time parameters can be achieved. Real time embedded authorization makes sharing of information more secure and facilitate the user through automated artificial intelligence and implementation in practical application. Moving towards the social impacts of this project, Facial recognition help the society to cope up with security issues as it is the only biometric that allows you to perform passive identification (e.g. Identifying a terrorist in busy Airport terminal). Last but not the least, the main target is that the device can easily be implemented in our daily life, its cost and size should be industry friendly. From using this project, facial recognition in smart security cameras especially in areas of surveillance, where security is the prime ask to acquire, such systems need to be implemented to have a proper check and balance.

Project Objectives

The common system of face recognition almost takes personal computer as development platform. In our project we are converting PC based system into an embedded system with fast processing speed, small volumeand price concessions so that the product begin to step into the range of people's horizons. Our project on real time embedded authorization makes sharing of information more secure and facilitate the user through automated artificial intelligence and achieving the goal to be implemented in real time application. The objectives are: 

• Increasing processing speed
• Converting PC based system into embedded system
• Feasibility of the system
• Attain maximum accuracy

Increasing processing speed

First objective of this project is to modify the existing face recognition system which are lagging from various latencies and throughputs  in terms of transient response and delay on large amount of data base, by upgrading existing system and implementing into surveillance areas.

Converting PC based system into embedded system

Secondly, to come up with a scalable modular hardware solution for real time face recognition on a large data base with minimum time . Our project consist of an embedded system and a camera. It works as a standalone device without personal computers.

Feasibility for the system

Another objective is to control the FPGA Controller with simple and efficient graphical environment of LabVIEW which does not use tons of lines of Verilog and VHDL language that are too complex to control the system design accessed through FPGA.

Attain maximum accuracy

Last but not the least, it is our prime objective to obtain maximum accuracy keeping all parameters under control by overcoming following issues:

• Environmental (resolution, illumination and dust)
• Synchronization errors (interface between camera and processor)
• Algorithm issues (complexity)

Project Implementation Method

The project implementation method is divided into four main domains.

Data Acquisition

The first step deals with acquiring data i.e. images, videos without any noise or environmental issues. The algorithm and the database is stored in the embedded system (MyRIO) and the embedded system is interface with the camera, the camera is working as an acquisition device which takes the input. The fps, resolution, illumination effects, dust effects and other environmental effects can be attenuates so that the embedded system can give efficient output. There are features given to user where the accuracy is more important so the user can increase the resolution, where time efficiency (no delay) is required so user can increase fps or it can be any situation.

Interfacing

The second step involves the process of interfacing, after the setting of modes the input images in the form of data is transferred to MyRIO (embedded system). The interface between MyRIO and camera in our project is a USB cable. The cable which we are using is USB 2.0 and USB 3.0. We synchronize the camera with the device(embedded system) so the interfacing delay is very low or negligible.

Image Preprocessing

Finally the embedded system takes the input with enhanced quality of image and removal of noise effects. We used some strategies to filter out the unwanted the impurities and to improve the resolution of image i.e. the use of histogram improves the brightness of Image by providing the equal parts to whole image rather than focusing one part of image. 

Applying Algorithms.

Algorithms that we are using detects the existence of human’s face, the algorithm is Camshift algorithm with Haar classifier. Camshaft detects the face area through maximum pixel density. It can detects the face when the object(person’s face) is near or far from camera and also with a rotation. The algorithm uses Haar classifier which is a machine learning based approach and match the feature with images in the database.

Removal of delay factor

The face classifier finishes face recognition by contrasting face with the feature extracted from image. Even a 24x24 window results over 160000 features and among all these features extracted by the haar classifier, some of them are irrelevant which also makes the system to take time. MyRIO is a parallel processor, additionally we are using parallel buffer which doubles the processing speed and reduce the effect of delay.

Benefits of the Project

Key Benefits

• It operates without user cooperation.
• It is the fastest face recognition technology with no delays and have transient response.
• It can be used in Security systems within high surveillance areas.
• Adoptable with concerned optimal response
• Flexible according to the user’s demand

Biometric Verification

Compared to other biometric applications, facial recognition is the more successful and the project take face recognition a step ahead in automation because the software is bassed on machine learning so it impoves itself after a particular time.

Social Benefits

The project, Facial recognition help the society to cope up with the security issues as this it’s the only biometric that allows you to perform passive identification (e.g. Identifying a terrorist in busy Airport terminal).

Reducing Processing time

In our project we are dividing the database in to two virtual buffers. So that we can process the complete database in half time.

Database handling Efficiency

It can be used by law enforcement agencies to detect the images and match them with the database. The system in our project can measure and match unique facial characteristics and accordingly identify and authenticate the person’s identity stored in the database. Our project isn’t just for apps and social media it can also be used with database. Now when working with database the two main conditions to keep in mind is the processing speed and memory handling capacity. In our project, we are using the hardware that is capable to process a large database with a high speed to cope up with the latencies issues.

Bringing Real time environment in surveillance Areas

Though before this other techniques were used, but couldn’t get off from nail of “delay factor”. From using this project, facial recognition in smart security cameras to its uses in digital medical applications, facial recognition software might help us in creating a safer, healthier future. Especially in areas of surveillance, where security is the prime ask to acquire, such system need to be implemented to have a proper check and balance.

Future Scope

Being a general purpose project, it has flexibility that it can be used anywhere, as it’s highly recommended for those areas or organizations which ask for tight security and cutting edge applications. So the customer can be estimated on the basis of number of organizations that are in the need of this tremendous design. 4 to 5 organizations (100 customers) we are estimating at the beginning.

Technical Details of Final Deliverable

The project consist on two main components that are MyRIO (embedded system) and a camera. The camera is working as an acquisition device which takes input in the form of images and video and apply preprocessing techniques to enhance the quality of image for better output results. The other component is the embedded system which handles the database and the programs that contains algorithms and other libraries.

ROLE OF AQUISATION DEVICE

With the aid of image acquisition device(the camera) the acquire images, after a series of preprocessing (as shown in the figure) which aims to enhance the precision of image and prepare for the better basis of feature extraction and recognition, are passed to the embedded system.

IMPORTANCE OF ALGORITHM

Algorithms that we are using detects the existence of human’s face, the algorithm we are using are Camshift algorithm with Haar classifier. Camshaft detects the face area through maximum pixel density. Camshift can detects the face when the object(person) is near to the camera or at a distance and with a rotation. The algorithm uses Haar classifier which is a machine learning based approach.The classifier finishes face recognition by contrasting face with the feature extracted from image. Even a 24x24 window(smallest image) results over 160000 features and among all these features extracted by the haar classifier, some of them are irrelevant which also makes the system to take time.

DESIGN DETAILS

The complete design is divided into two parts, face static recognition and dynamic tracking. We extract the features of face from image and use algorithm in a template matching to calculate the degree of similarity between image and template and in the other design of face dynamic tracking, we adopt the method that combines Haar classifier with Camshift algorithm. The dynamic tracking takes time to process the complete algorithm on database. The algorithm gives the accuracy of 96.33% and it takes almost 253 msec.

TIME EFFICIENCY OF THE SYSTEM

To reduce this delay and get transient response with a database, we use an embedded system which is designed for real time application, MyRIO. MyRIO has built-in FPGA board which enables it to do parallel processing and have a sufficient storage which reduces the time a processor requires during the interfacing with an SD card. MyRIO have latest technology of ZINC IC and its another version is using in industry with the name CompactRIO.

Final Deliverable of the Project HW/SW integrated systemType of Industry Education , Legal , Media , Security , Telecommunication Technologies Artificial Intelligence(AI), Robotics, NeuroTech, Wearables and ImplantablesSustainable Development Goals Industry, Innovation and Infrastructure, Sustainable Cities and Communities, Responsible Consumption and Production, Peace and Justice Strong InstitutionsRequired Resources

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