Rahiddin, Rahillda Nadhirah Norizzaty (2025) Texture representation for wood defect image classification using enhanced colour uniform local binary pattern. Doctoral thesis, Universiti Teknikal Malaysia Melaka.
![[img]](http://eprints.utem.edu.my/style/images/fileicons/text.png) |
Text (24 Pages)
Texture representation for wood defect image classification using enhanced colour uniform local binary pattern (24 pages).pdf - Submitted Version Download (961kB) |
|
Text (Full Text)
Texture representation for wood defect image classification using enhanced colour uniform local binary pattern.pdf - Submitted Version Restricted to Registered users only Download (2MB) |
Abstract
Wood defect classification remains a critical challenge in the timber industry due to natural variations in defect appearance, illumination inconsistencies, and texture complexity across wood species. Traditional manual inspection methods for assessing wood quality are time-consuming, subjective, and prone to human error, limiting their reliability for large-scale industrial applications. While computational approaches have been developed to automate defect detection, existing methods struggle with distinguishing defect classes due to grayscale-based texture extraction, sensitivity to lighting conditions, and excessive feature generation that leads to overfitting. Previous research has primarily relied on Basic Local Binary Patterns (LBP) to extract surface texture features, treating defects as localized structural variations. However, LBP operates on grayscale images, resulting in the loss of colour-based information, which is essential for distinguishing defects from clear wood regions. Additionally, LBP exhibits sensitivity to illumination changes, meaning classification performance varies under different lighting conditions, leading to inconsistencies in defect identification. High-resolution images introduce further complexity, as LBP generates excessive features, increasing the risk of overfitting and poor generalization in classification models. This study aims to address these limitations by proposing an enhanced feature representation that utilizes Colour Uniform Local Binary Patterns (CULBP) combined with comprehensive colour normalization to mitigate the effects of varying lighting conditions and improve classification robustness. The methodology extracts texture and colour features from images of four wood species which are Rubberwood, Kembang Semangkuk (KSK), Merbau, and Meranti and evaluates classification accuracy across multiple defect types, including bark pocket, blue stain, borer holes, brown stain, knot, rot, split, and wane. A succession of colour feature sets was derived from individual and combined RGB channels, with classification performed using an Artificial Neural Network (ANN). Results indicate that RGB channels, without substantial colour normalization, achieve the highest classification accuracy at 90.2%, surpassing previous grayscale-based methods such as Daubechies Wavelet and Basic LBP (85%) and Basic LBP alone (65.4%). Among the evaluated species, Rubberwood exhibited the highest accuracy, with particularly strong results for Rubberwood Rot (92.2%) and Merbau Hole (95.6%), demonstrating the advantages of colour-based feature extraction in defect classification. By integrating multi-channel texture and colour features, this study provides a more robust defect classification framework, addressing key challenges related to illumination sensitivity, texture inconsistencies, and excessive feature extraction. The findings highlight the importance of systematic preprocessing using colour normalization and statistical validation using Multivariate Analysis of Variance (MANOVA) to improve classification performance. These advancements offer a scalable solution for industrial wood quality assessment, contributing to enhanced defect recognition and efficiency in the timber industry.
| Item Type: | Thesis (Doctoral) |
|---|---|
| Uncontrolled Keywords: | Feature extraction, Local binary pattern, Wood defects, Artificial neural network, Colour uniform local binary pattern |
| Subjects: | T Technology T Technology > TA Engineering (General). Civil engineering (General) |
| Divisions: | Faculty of Information and Communication Technology |
| Depositing User: | Norhairol Khalid |
| Date Deposited: | 21 Jan 2026 07:21 |
| Last Modified: | 21 Jan 2026 07:21 |
| URI: | http://eprints.utem.edu.my/id/eprint/29453 |
| Statistic Details: | View Download Statistic |
Actions (login required)
 |
View Item |