Distributed System on Chip Design for Face Recognition

Project Title

Distributed System on Chip Design for Face Recognition

Project Area of Specialization Cyber SecurityProject Summary

The traditional closed-circuit television (CCTV) system needs a human 24/7 for monitoring which is costly and insufficient. The automatic recognition of faces in CCTV images can help many organizations such as the law-enforcement to identify the suspect, missing person, and a person entering a restricted territory. However, an image-based recognition has many issues such as scaling, rotation, cluttered background, and variation in light intensity. In this project, we aims to develop CCTV image) based human face recognition system using convolutional neural network (CNN). The proposed system will be implemented in distributed paradigm. The first component of the system includes an image acquisitioning from CCTV, preprocessing of image, detection and localization of faces in an image, extraction of faces from image, and network infrastructure to send the image to recognition server for recognizing the person. All the data is encrypted. The second component includes a recognition server using CNN trained with more than 40k images. Finally, the project goals are to detect multiple faces in an image and recognize them in minimum computing time and accuracy more than 90%.

Project Objectives

1. Detection of Face of the person in an image
2. CNN based trend platform for face recognition
3. Raspberry-pi based system for face detection
4. Transmission of images to Recognition system
5. Encryption Decryption of transmitted face image
6. 40k CCTV image Database collection
7. Minimum 90% accuracy for recognition

Project Implementation Method

The whole architecture of the project is shown below, where the input data passes from different phases. First phase is acquisition phase. In acquisition phase we acquire an image. Images need to be restored from the source (usually a hardware source) image sensor, making it the first step in the workflow sequence because processing is not possible. Our system on chip constantly reads images, which is our pre-processed input. For image enhancementweimproved low contrast, color or low light image. Image preprocessing and standardization are an integral part of the face recognition system. For the best results, we applied a canny filter. After applying this filter, we found a sharp image with edges. At face detection stage, we detect the face in the entire image. That image will be used for face recognition. This image contains only the faces detected from a specific algorithm.Our region of interest (ROI) is face. In face we are focusing on eyes, eye brow, nose, forehead and lips. We resize the image to a specific size and then we will send it to the training phase. When we get a ROI through system on chip, we first extract the features and then save it to the database for identification. Recognition is the final step in our process, comparing the received image to the database, and if it has the same characteristics as in the database.

Below is the flow chart of whole system,

Benefits of the Project

1. There will be no need for a human to monitor people. As monitoring by human from CCTV camera involves problems in reliability and scalability as human can’t recognizes every person. So the project will be fully automated
2. This project has aim to improve the security level
3. This project will provide high accuracy rates of recognition. So it can give criminals hard time
4.  The security staff will not favor any other employee in attendance. The project will note time when person enters.
5.  Improves Convenience for Everyday Life
6.  All the records will be maintained

Technical Details of Final Deliverable

• CCTV and Raspberry-PI based system to acquire the images
• Development of algorithm to pre-process to enhance the image, remove noise, and acceptable for efficient recognition
• Algorithm for Face detection in CCTV image
• System will be secured using encryption techniques
• Convolutional Neural Network based platform for face recognition

Final Deliverable of the Project HW/SW integrated systemType of Industry IT , Security , Telecommunication Technologies Artificial Intelligence(AI), Internet of Things (IoT)Sustainable Development Goals Peace and Justice Strong InstitutionsRequired Resources

Item Name	Type	No. of Units	Per Unit Cost (in Rs)	Total (in Rs)
			Total in (Rs)	80000
Image Sensor	Equipment	2	18000	36000
System on chip (Raspberry-pi) with kit	Equipment	2	15000	30000
Wi-Fi Module	Equipment	2	2000	4000
Printing of thesis copies	Miscellaneous	1	10000	10000