Drunk Classification with Infrared Face Image



Goh, Choo Lim (2021) Drunk Classification with Infrared Face Image. Final Year Project (Bachelor), Tunku Abdul Rahman University College.

[img] Text
Goh Choo Lim.pdf
Restricted to Registered users only

Download (3MB)


In this work, a proposed Convolutional Neural Network is used to perform drunk discrimination of person by its thermal face images. The proposed CNN architecture is performed and modified to have a better drunk discrimination performance after the discovering and comparison had done on the literature review stages. The thermal camera working principles on thermal capture and the thermoregulation effect on the human body is the primary focus of this work to classify drunk. During the implementation state, the modeling and simulation environment of the system is validated using the dataset sample, the SOBER-DRUNK provided by Georgia Koukiou and Vassillis Anastassopoulos created in the Electronics Laboratory, Greece. A Convolutional Neural Network architecture with modified layers is built to perform drunk discrimination. The drunk samples are trained under the Convolutional Neural Network created by the software, Matlab 2020b with Image Processing Library and the result are examined. A comparison is performed on the proposed Convolutional Neural Network architecture and the Basic Convolutional Neural Network in terms of accuracy and training time consumption. The performance of the proposed Convolutional Neural Network is discussed. The accuracy of the proposed Convolutional Neural Network had achieved 85% and the time consumption required is reduced by 25.9% as compared to the same data using Basic Convolutional Neural Network Architecture.

Item Type: Final Year Project
Subjects: Technology > Electrical engineering. Electronics engineering
Faculties: Faculty of Engineering and Technology > Bachelor of Electrical and Electronics Engineering with Honours
Depositing User: Library Staff
Date Deposited: 09 Jul 2021 07:55
Last Modified: 09 Jul 2021 07:55
URI: https://eprints.tarc.edu.my/id/eprint/18656