Automatic Face Verification using Raspberry Pi

Nowadays, Face Recognition (FR) is being extensively studied for past few decades in many real-life applications [1]. The FR is divided into two main categories, which are (i) Face Recognition, also known as 1:N match and (ii) Face Verification, which is also k

Project Title

Automatic Face Verification using Raspberry Pi

Project Area of Specialization

Artificial Intelligence

Project Summary

Nowadays, Face Recognition (FR) is being extensively studied for past few decades in many real-life applications [1]. The FR is divided into two main categories, which are (i) Face Recognition, also known as 1:N match and (ii) Face Verification, which is also known as 1:1 match [2],[3]. Recently the biometric based techniques, such as fingerprints recognition, voice recognition, palm, or veins have emerged as promising tools for access control, but these techniques require special arrangements and can expose individuals to germs. Compared with these biometrics, the human face has a better potential to categorize the identity in a safe and non-intrusive manner. A single camera can be installed at any public place to capture facial images. An automated Face Verification System will be helpful to the users without any health-related risks. Therefore, the FR is nowadays extensively used in many important application areas, such as surveillance, forensics, and security control [4].  Due to various challenging issues, such as pose variation, occlusion, low-resolution, and non-uniform illuminations, the FR is still a challenging task today [5]. Most of the current state-of-the-art, FR algorithms report high recognition accuracy on aforementioned issues and under constraint environments but struggle in uncontrolled environment [6]. The occlusion is one of the main hurdles that severely affects the recognition accuracy of most of the FR algorithms [1]. This project will focus to develop an automatic face verification system using Raspberry-Pi. Specifically, the proposed project aims to achieve high verification accuracy on real life frontal and occluded face images. The proposed FR algorithm will initially locate face in the input images. Later, the detected face will be verified using a robust verification algorithm. Finally, the verified face will be displayed on a screen. The developed system will be interfaced with Raspberry-Pi to show the final verified face from the stored faces database gallery.

Project Objectives

The objective of our project is to develop an efficient and highly accurate face verification system using Raspberry-Pi to verify the real time frontal and occluded face images. Specifically, the objectives of the proposed project are:

• To develop a user friendly and state-of-the-art face verification system using Raspberry-pi platform on frontal and occluded face images. We believe that such verification system will enable security staff to efficiently keep track of people at various locations.
• To develop an accurate, reliable, and robust face verification system on the aforementioned specifications that can verify faces in real-time.
• We focus on a user-friendly face verification system for the proposed project for the security and inspection staff, who may have visual, physical, or cognitive issues.

Project Implementation Method

To begin this project, we will take start from understandings of complex machine learning algorithms of face detection and face verification [1],[2]. Then feasibility of Raspberry-pi will be explored in depth. The schematic design of the proposed Face verification system is depicted in Figure 1. The supervisor and the project team members are competent enough to successfully complete all of the tasks. Recently, the project team members have successfully developed certain object detection and recognition applications [3], [4], and [5]. Moreover, the supervisor has vast experience in image analysis features extraction, recognition, and image understanding [6], [7], [8], [9],[10]. The experience of the supervisor will be very helpful to design and develop an automated face verification system. We believe in the happy and safe living life of all humans. Therefore, to develop such a human friendly system is one of the major motivations to proceed to the proposed work. Face detection, features extraction, and face verification operations when integrated with the Raspberry-pi will lead to a high level of accuracy of the system due to the enormous computation and memory resources. Because, these operations involve complex and intensive computer vision, machine learning, and image processing algorithms that are executed in high dimensional space. These complex computations are one of the bottlenecks in development of more accurate object detection and recognition solutions.

References

1. Z. Mahmood, N. Muhammad, N. Bibi, and T. Ali, “A review on state-of-the-art face recognition approaches,” Fractals, Complex Geometry Patterns and Scaling in Nature and Society, Vol. 25, No. 2, 2017, pp. 1750025-1-1750025-19.
2. M. Haq, A. Shahzad, Z. Mahmood, A. Shah, N. Muhammad, and T. Akram, “Boosting the Face Recognition Performance of Ensemble Based LDA for Pose, Non-uniform Illuminations, and Low-Resolution Images,” KSII Transactions on Internet and Information Systems, 2019, Vol. 13, No. 6, pp. 3144-3164.
3. K. Fareed, F. Sultan, K. Khan, and Z. Mahmood, “A Robust Face Recognition Method for Expression and Pose Variant Images,” 14th International Conference on Open Source Systems and Technologies (ICOSST), 2020. (Accepted and to appear).
4. Z. Mahmood, N. Bibi, M. Usman, U. Khan, and N. Muhammad, “Mobile cloud based framework for sports applications,” Multidimensional Systems and Signal Processing, Vol. 30, No. 4, 2019, pp. 1991?2019.
5. Z. Mahmood, O. Haneef, N. Muhammad, and S. Khattak, “Towards a fully automated car parking system,” IET Intelligent Transport Systems, 2018, Vol. 13, No. 2, pp. 293?302.
6. Z. Mahmood, T. Ali, N. Muhammad, N. Bibi, I. Shahzad, and S. Azmat, “EAR: Enhanced augmented reality system for sports entertainment applications,” KSII Transactions on Internet and Information Systems, Vol. 11, No. 12, 2017, pp. 6069?6091.

Benefits of the Project

Due to the development of the proposed project, many organizations in Pakistan can take the following benefits:

• The proposed project will provide suggestions based on the verification results of faces from available data. In addition, the proposed project will provide improved public security.
• The proposed project will reduce the unnecessary and time-consuming process for manual security checking. It will provide fast and non-intrusive identity verification mechanism.
• The proposed project will act as an online security guard that has more capability than security personnel (human). It will assist the crime investigation and forensic evidence evaluations quickly.
• The proposed project will reduce the unnecessary and time-consuming human interaction with the system. The proposed system can be used in universities. We believe it will provide better attendance record. It can also be use in other organizations, such as banking where it will reduce online banking frauds.
• The proposed project will provide high order thinking, decision power, stability, availability, and security to the system.
• We believe that the proposed project is proactive, responsive, and works in real time environment, due to a fact that newly hired staff with less experience can operate this product. Importantly, there is no need to train the IT personnel.

 References

7.       Z. Mahmood, T. Ali, and S. U. Khan, “Effects of pose and image resolution on automatic face recognition,” IET Biometrics, 2016, Vol. 5, No. 2, pp. 111?119.

8.       Z. Mahmood, T. Ali, S. Khattak, L. Hassan, and S. U. Khan “Automatic player detection and identification for sports entertainment applications,” Pattern Analysis and Applications, 2015, Vol. 18, No. 4, pp. 971?982.

9.       H. Ullah, M. Haq, S. Khattak, G. Z. Khan, and Z. Mahmood, “A Robust Face Recognition Method for Occluded and Low-Resolution Images,” 2nd International Conference on Applied and Engineering Mathematics (ICAEM), 2019, pp. 86?91.

10.     O. Haneef, S. Maqbool, F. Siddique, Z. Mahmood, S. Khattak, and G. Z. Khan, “Effects of Image Resolution on Automatic Face Detection,” 2nd International Conference on Communication, Computing and Digital Systems (C-CODE), 2019, pp. 231?236.

Technical Details of Final Deliverable

The flow of the proposed face verification system is depicted in Fig. 1. The system will comprise of the following four modules:

Step-1: Image Acquisition: The first step in expert systems involves the acquisition of the face. In our work, the face image will be acquired through a camera embedded in Raspberry-pi. Our initial study reveals that Raspberry-pi module V2 contains 8 Mega-pixels camera, which is sufficient for our system development.

Step-2: Object Detection: In the proposed project, object detection will be studied and investigated in the context of human face. To accurately detect the human face, a robust methodology will be explored, such as deep learning and machine learning methods [2],[3]. The ultimate face region will be spotted by automatically drawing a rectangle/circle around the face, which is commonly called bounding box. For face verification, only space inside the bounding box will be analyzed to extract and identify features to conclude its identity.

Step-3: Features Extraction: Features are the visual cues that will be used for face verification. In our work, face features will be extracted through an automatic classification error driven mechanism that will be deployed in verification stage [4]. The extracted features will be measurable with high sensitivity. The extracted features will also have high correlation with face features, which is high probability of true positive response. Furthermore, the extracted features will have high specificity, which is high probability of true negative response. We will explore state-of-the-art feature extractors, for instance [2],[5] that will effectively help to detect human face.

Step-4: Object Verification: In the proposed project, object verification will be studied and investigated in the context of face based on the outcome of the Step-2 and Step 3. To verify a human face, a feature comparison and recognition module will be deployed at the end [5]. In recent times, large numbers of published works have appeared in literatures that focus on face recognition [6]. However, aforementioned schemes process images from standard databases and many times require special arrangements to develop the face image database. Moreover, the computation time is a crucial factor during the algorithm development. However, in our proposed project, we will apply a boosted and modified version of the Local Binary Pattern (LBP) based approach in the complex face verification task [7]. We believe that the proposed verification scheme will be able to perform recognition task in higher dimension space that will result in robustness and improved accuracy. The feature classification module at the end will produce a score that will indicate the status of face, which is either verified or unverified. Based on the outcome of the verification module, a suitable action can be recommended to the authorities.

Fig 1: Architecture of proposed Face Verification System

Final Deliverable of the Project

HW/SW integrated system

Core Industry

IT

Other Industries

Telecommunication

Core Technology

Artificial Intelligence(AI)

Other Technologies

Robotics

Sustainable Development Goals

Industry, Innovation and Infrastructure

Required Resources

If you need this project, please contact me on contact@adikhanofficial.com