A Study of Facial Recognition for Crowded Indoor Images

Lim, Jing Wei (2016) A Study of Facial Recognition for Crowded Indoor Images. Final Year Project (Bachelor), Tunku Abdul Rahman University College.

Text Lim Jing Wei.pdf Restricted to Registered users only Download (3MB)

Abstract

Every lecturer will certainly encounter a problem of having difficulty in taking attendance of all students present in a lecture hall and this problem is pronounced when the lecturer is conducting a lecture with students of over 100 pax. More often, the lecturer would either spend approximately 5 to 10 minutes to take their attendance or circulate the attendance list around to let students do the attendance checks themselves. However, the abovementioned two approaches have their own underlying shortcomings, such as the former wastes time to perform attendance checks by lecturers themselves while the students might falsify their attendance records for the latter one. On the other hand, facial recognition system is a type of computer application that can recognize persons from images or videos by analysing and comparing the facial patterns enrolled in its database since every face carries specific and unique facial features that can be a distinctive trait. It becomes prevalent in our daily life thanks to the technologies of both image processing and machine learning advance as time progresses. Prior to facial recognition, the system has to first find out which one is face by adapting facial detection and alignment module from image processing as well as trained with the subjects’ faces using machine learning that emulates how the brain wires and how the vision receptive fields behave in a human visual perception. . In the project, the facial recognition engine is trained with frontal facial images of students by means of machine learning beforehand. Subsequently, the image processing is tasked to detect faces present in the image and crop them to feed into facial recognition system to perform facial recognition. As a result, the recognized faces and respective confidence rates are shown. Henceforth the problem stated in the former paragraph can be alleviated by implementing facial recognition technique in recognizing faces in a crowded indoor images.