UUM Electronic Theses and Dissertation
UUM ETD | Universiti Utara Malaysian Electronic Theses and Dissertation
FAQs | Feedback | Search Tips | Sitemap

Classification Modeling for Malaysian Blooming Flower Images Using Neural Networks

Muhammad Ashraq, Salahuddin (2013) Classification Modeling for Malaysian Blooming Flower Images Using Neural Networks. Masters thesis, Universiti Utara Malaysia.

[thumbnail of s803116.pdf]
Preview
Text
s803116.pdf

Download (3MB) | Preview
[thumbnail of s803116.pdf]
Preview
Text
s803116.pdf

Download (502kB) | Preview

Abstract

Image processing is a rapidly growing research area of computer science and remains as a challenging problem within the computer vision fields. For the classification of flower images, the problem is mainly due to the huge similarities in terms of colour and texture. The appearance of the image itself such as variation of lights due to different lighting condition, shadow effect on the object’s surface, size, shape, rotation and position, background clutter, states of blooming or budding may affect the utilized classification techniques. This study aims to develop a classification model for Malaysian blooming flowers using neural network with the back propagation algorithms. The flower image is extracted through Region of Interest (ROI) in which texture and colour are emphasized in this study. In this research, a total of 960 images were extracted from 16 types of flowers. Each ROI was represented by three colour attributes (Hue, Saturation, and Value) and four textures attribute (Contrast, Correlation, Energy and Homogeneity). In training and testing phases, experiments were carried out to observe the classification performance of Neural Networks with duplication of difficult pattern to learn (referred to as DOUBLE) as this could possibly explain as to why some flower images were difficult to learn by classifiers. Results show that the overall performance of Neural Network with DOUBLE is 96.3% while actual data set is 68.3%, and the accuracy obtained from Logistic Regression with actual data set is 60.5%. The Decision Tree classification results indicate that the highest performance obtained by Chi-Squared Automatic Interaction Detection(CHAID) and Exhaustive CHAID (EX-CHAID) is merely 42% with DOUBLE. The findings from this study indicate that Neural Network with DOUBLE data set produces highest performance compared to Logistic Regression and Decision Tree. Therefore, NN has been potential in building Malaysian blooming flower model. Future studies can be focused on increasing the sample size and ROI thus may lead to a higher percentage of accuracy. Nevertheless, the developed flower model can be used as part of the Malaysian Blooming Flower recognition system in the future where the colours and texture are needed in the flower identification process.

Item Type: Thesis (Masters)
Supervisor : Siraj, Fadzilah and Zulkifli, Abdul Nasir
Item ID: 3847
Uncontrolled Keywords: Image Classification, Neural Networks, Multilayer Perceptron, Decision Tree, Logistic Regression
Subjects: Q Science > QA Mathematics
Divisions: Awang Had Salleh Graduate School of Arts & Sciences
Date Deposited: 09 Mar 2014 00:33
Last Modified: 18 Jan 2023 08:32
Department: Awang Had Salleh Graduate School of Arts and Sciences
Name: Siraj, Fadzilah and Zulkifli, Abdul Nasir
URI: https://etd.uum.edu.my/id/eprint/3847

Actions (login required)

View Item
View Item