Face Recognition For Drunk Driver

Muhammad Fakhrul Alif, Heirman (2023) Face Recognition For Drunk Driver. [Final Year Project Report] (Unpublished)

	PDF MUHAMMAD FAKHRUL ALIF BIN HEIRMAN (24pgs).pdf Download (711kB)
	PDF (Please get the password by email to repository@unimas.my , or call ext: 3914 / 3942 / 3933) MUHAMMAD FAKHRUL ALIF BIN HEIRMAN (Fulltext).pdf Restricted to Registered users only Download (3MB)

Abstract

This project's objective is to determine whether there is a discernible difference between the face behaviours of persons who are drunk and those of people who are sober. There is a significant amount of face recognition that occurs in situations involving intoxicated persons. It is imperative that intoxicated drivers be identified as soon as possible, preferably before they get behind the wheel, to reduce the likelihood that an accident will take place. At the outset, a comparison is made between the facial behaviour of a drunk person and that of a sober person. This is followed by the collection of comprehensive data regarding the distinction between the two face behaviours through the utilisation of the data that is produced following the training of the digital images through Image Processing. This procedure has the potential to demonstrate that the distinction between intoxicated and sober people can be identified based on facial behaviour, and that the distinction can be spotted using computer vision. The first approach for achieving the project's objectives is to run the face detection method, which detects the red colour on the eyes, nose, and cheeks by utilising Haar Cascade to identify just the face and obtain the confidence rate value. The project proceeds by completing the training and testing processes for all digital photographs, which are classified into three cases. A total of 100 digital images of sober face and 100 digital images of drunk face from single individuals for case 1. For case 2, 150 digital images of drunk face and 150 digital images of sober face from multiple individuals. As for case 3, a total of 50 digital images of sober faces and 50 digital images of drunk faces from multiple individuals. To meet the project's goals, a probability comparison from the testing procedure between drunk and sober is performed. The confidence rate of red colour detection for eyes, nose and cheek is between 75% until 98% for drunk face and 35% until 74% for sober face. Meanwhile, the probability of drunk face is approximately between 0.94 until 0.99 and probability of sober face is between 0.62 until 0.84 for case 1. Next, the probability of drunk face is between 0.79 until 0.95 and the probability for sober face is between 0.78 until 1.00 for case 2. For case 3, the probability of drunk face is between 0.62 until 0.99 and the probability of between 0.91 until 0.99 was for sober face.

Actions (For repository members only: login required)

View Item