PCA Method in Facial Features Extraction for Attendance Tracking Prototype
Farhan Najmi Mohd Nazri, Siti Nur Kamaliah Kamarudin, Mohd Razif Shamsuddin · 2022
Facial recognition has long become a way to identify an individual-it represents an identity of a person. For this study, an automated students' attendance tracking prototype based on facial recognition was proposed, and later developed. Facial recognition systems can be extremely helpful in real-world applications, particularly in security control systems. For this study, firstly video framing was prepared by activating the camera via a user-friendly interface. Using the Multi-Task Cascaded Convolutional Neural Network (MTCNN) algorithm, the face Region of Interest (ROI) was then identified, and segmented from the obtained video frames. In order to minimize loss of information, the images were resized, cropped, and RGB-greyed at the pre-processing step. Principal Component Analysis (PCA) was then used to extract the characteristics from facial images during the face recognition stage. The extracted features from the test images were then compared to the extracted features from the training images. The facial images were then categorised, and identified using the PCA algorithm. Finally, the identified student's attendance will be recorded for tracking purposes, and stored in a local database. One of the main advantages of the prototype is the automated registration of new students where Adhoc registration are allowed using the prototype. The study managed to achieve the best recognition accuracy of 98.7%. For future works, it is suggested that a higher number of images per individual to be used, with different pre-processing techniques.