Study with the several resources on Docsity
Earn points by helping other students or get them with a premium plan
Prepare for your exams
Study with the several resources on Docsity
Earn points to download
Earn points by helping other students or get them with a premium plan
management systemmanagement systemmanagement systemmanagement systemmanagement system
Typology: Thesis
1 / 103
This page cannot be seen from the preview
Don't miss anything!
UNIVERSITY OF KWAZULU-NATAL, COLLEGE OF AGRICULTURE, ENGINEERING AND SCIENCE DECLARATION
The research described in this thesis was performed at the University of KwaZulu-Natal under the supervision of Prof. Jules-Raymond Tapamo and. I hereby declare that all materials incorporated in this thesis are my own original work except where the acknowledgement is made by name or in the form of reference. The work contained herein has not been submitted in part or whole for a degree at any other university.
Signed:................................................... Name: Louis Mothwa Date: August 2019
As the candidate’s supervisor, I have approved this thesis for submission.
Signed:................................................... Name: Prof. Jules-Raymond Tapamo Date: August 2019
As the candidate’s co-supervisor, I have approved this thesis for submission.
Signed:....................................................... Name: Dr Temitope Mapayi Date: August 2019
i
UNIVERSITY OF KWAZULU-NATAL, COLLEGE OF AGRICULTURE, ENGINEERING AND SCIENCE DECLARATION 2 - PUBLICATIONS
I, Louis Mothwa, declare that the following publications from part of this dissertation.
Signed:..............................................................
iii
Acknowledgements
First I would like to thank God for giving me strength, wisdom, and understanding throughout my work.
This work was supported and funded by the Armaments Corporation of South Africa (Armsco). I am grateful to have this opportunity to fulfill my dreams.
I thank my Supervisors Prof J.R. Tapamo and Dr. T. Mapayi for their support, dedication, patience, and endless insight. This feat could not have been possible without your assistance. I am grateful to the University of KwaZulu-Natal students who offered to appear in the live experiments conducted to collect data to produce training and testing dataset’s used in this dissertation.
Lastly, I would like to thank my family and friends, for their emotional support and endless motivation, which encouraged me to be determined till the end.
iv
Declaration of Authorship i
List of Publications iii
Acknowledgements iv
Abstract v
Contents vi List of Figures viii
List of Tables x List of Algorithms xi
List of Acronyms xii
1 Introduction 1 1.1 Background....................................... 1 1.2 Overview of Computer Vision Systems and Applications.............. 3 1.2.1 Image Acquisition............................... 3 1.2.2 Image Processing................................ 4 1.2.3 Images Analysis................................. 4 1.3 Motivation........................................ 4 1.4 Problem Statement................................... 5 1.5 Research Aim and Objectives............................. 6 1.6 Scope Of The Study And Limitations......................... 6 1.7 Contributions...................................... 7 1.8 Overview........................................ 7 2 Literature Review 8 2.1 Introduction....................................... 8 2.2 Attendance Monitoring System Technologies..................... 9 2.2.1 RFID Authentication System......................... 9 2.2.2 Mobile Smart Attendance System....................... 9 2.2.3 Bluetooth Based Attendance Monitoring System.............. 10 2.2.4 Fingerprint Attendance Monitoring System................. 10 2.2.5 Attendance Monitoring Systems Using Face Recognition.......... 12 2.2.6 Multiple-Camera Positioning And Collaboration.............. 15 2.3 Face Recognition.................................... 17 2.3.1 Face Detection................................. 17 2.3.2 Image Pre-Processing............................. 20 2.3.3 Feature Extraction And Classification.................... 21 2.3.4 Global Feature Extraction And Classification Methods........... 21 2.3.5 Local Feature Extraction And Classification Methods............ 23 vi
List of Figures ix 4.6 Live scenarios captured from the UKZN classroom. The scenes shows the im-
5.1 Face detection rate from differently positioned cameras at a fixed time...... 65 5.2 The trained results of the BPNN with different number of neurons in the middle layer........................................... 70 5.3 Learning rate against error rate of the BPNN.................... 71 5.4 Comparison of the performance of the proposed method to existing methods using similar feature extraction and classification algorithms............... 72 5.5 Example of register of students attendance for the whole lecture period...... 73
x
AMS Attendance Monitoring System
ANN Artificial Neural Network
CVS Computer Vision System
ED Euclidean Distance
ICA Independent Component Analysis
LDA Linear Discriminant Analysis
LBP Local Binary Patterns
NB Naive Bayes
PCA Principle Component Analysis
RFID Radio Frequency Identification
SVM Support Vector Machine
xii
1.1 Background
An Attendance Monitoring System Attendance Monitoring System (AMS ) validates whether a person has attended a meeting/lecture or not. Attendance monitoring system can be used to confirm that the right people are in the right place at the right time [14].
Recent studies have shown that there is a correlation between student’s academic performance and punctuality to classes [41, 27]. Therefore, academic institutions need to be equipped with a proper way of managing attendance in classrooms and examination venues.
Attendance management can be carried out in many ways. This includes lecturers invigilating students, circulation of attendance register for students to indicate their presence by appending their signature, or automated attendance management systems. However, many attendance monitoring systems are prone to problems such as, false results, time wasting, cheat, and use of expensive materials [87]. As a result computer vision is being explored to build a better attendance management systems.
Humans are visual creatures [102]. Almost every human depends on sight to interpret the world. Most information and discoveries are products of the visual faculty. The interpretation of visuals makes the world to be informative and advances intelligence. Although sight is not the only chan- nel to acquire information, it is regarded as the greatest source for collecting information and grow
Introduction 3
can enable the processing of huge amount of data faster, can generate safe storage, and can be made more appropriate to real-time [102].
1.2 Overview of Computer Vision Systems and Applications
The major building blocks of computer vision applications are image acquisition, image processing and image analysis. Most computer vision systems are built from this protocol. Figure 1. illustrate the Computer Vision System (CVS) where the training and recognition process are shown.
Figure 1.2: Generic architecture of computer vision system using face recognition
Image acquisition is the process of locating and capturing the desired image. In the case of class attendance monitoring this process will further to search faces in a given frame. This process is mainly focused on locating the object of interest. In this project images will be captured using cameras and converted into a grayscale image for further processing and analysis [100, 113, 31, 106].
Introduction 4
The second phase of computer vision is image enhancement. Most computer vision algorithms such as computational photography and object recognition require data to be pre-processed in order to be enhance the phenomenon targeted in the image [102]. Images acquired from the first stage can contain some defects. Hence, the imperfections in the raw image have to be reduced for better analysis of the image. Images can be affected by exposure, color balancing, image noise, unbalanced light, rotations, low sharpness, and blurriness [100]. Defect can be caused by the quality of the camera or some environmental effects such as a drastic change in light. The major challenge is that most images captured always have noise that cannot be analyzed with naked eyes. Therefore, it is important to always enhance an image before processing [100, 9, 102].
Image analysis involves feature extraction and feature classification of a given data. In feature extraction, the best features representing an image are selected [9], while feature classification clusters homogeneous feature separately [102]. These features can be used to distinguish the images. Features can also be used for interpretation, classification, recognition and decision making between images. Object recognition performance heavily depends on feature extraction and classification algorithms used [100].
1.3 Motivation
Recent studies have shown that students attendance correlates with the academic improvements [53]. The moment students are in a class listening to a lecturer, writing notes, and participating on the lecturers lessons, leaves knowledge in their minds. Therefore, developing a reliable and accurate attendance monitoring system to record student’s presence through the entire lecture, could be of a great help to enhance the quality of academic deliveries.
The use of facial recognition has the potential to improve the automated attendance management systems. It is well documented that face recognition has improved significantly [85, 87, 68, 22, 52, 19]. One of the most important characteristics of facial recognition based attendance
Introduction 6
Existing automated modalities such as mobile application, RFID and Bluetooth based-systems are also prone to false results, intrinsic models, time wasting, and expensive instruments [21, 90]. Hence, the need to improve the attendance monitoring system is necessary. True authentication and identification are important features for attendance recording.
Although, facial recognition applications can be susceptible to false positives and false negatives, invariant lighting, facial occlusion, and drastic change in facial expression [87, 85, 68]. A well- modeled face recognition system, can improve the attendance management, due to its non-intrinsic nature [21]. This non-intrisic nature makes facial recognition biometric system to be essentially preferred when compared to other forms of biometrics.
1.5 Research Aim and Objectives
The main aim of this research is to design and implement an automated smart model that employs real-time face recognition to monitor student attendance without time wasting and physical interaction with the system. In order to achieve the main goal, the following specific objectives must be carried out:
1.6 Scope Of The Study And Limitations
This section elaborates the limitations of the investigations performed in this research.
Introduction 7
The video and images dataset of students in a classroom is very rare online. The proposed model was tested with a private database of students in a classroom.
1.7 Contributions
The main contributions of this research are:
1.8 Overview
Chapter 2 reviews the state-of-art of the facial recognition process. It elaborates on the contextu- alization of existing attendance management systems. Chapter 3 presents the system model and the Graphic User Interface (GUI) of the proposed model. In chapter 4, materials, algorithms, and methods such as face detection, image pre-processing, and feature extraction are presented. A full discussion on the experimental conditions, the results of the experiments from different classification algorithms, is presented in chapter 5. Chapter 6 draws the conclusion and presents recommendations for future work. Consent for the use and publication of images captured in the classroom is presented as appendix at the end of the dissertation.
