Profiling Wood Surfaces for Defect Detection Using Machine Vision and Image Processing

Lye, Chi Hin (2025) Profiling Wood Surfaces for Defect Detection Using Machine Vision and Image Processing. Final Year Project (Bachelor), Tunku Abdul Rahman University of Management and Technology.

Text Lye Chi Hin_Full Text.pdf Restricted to Registered users only Download (3MB)

Abstract

This research addresses the challenges faced by the wood frame industry in ensuring the quality of finished wood surfaces, particularly in the detection of defects such as cracks, dents, gesso line, and paint problems. One of the most challenging parts is to detect the gesso line defects on black-painted wood surfaces which requires good lighting conditions. The reliance on manual inspection methods is time-consuming, subjective and prone to human error. Therefore, the objective of this study is to develop and implement an automated defect detection system for wood surfaces using machine vision techniques. The methodology involves gathering comprehensive datasets of wood surface images containing various types of defects where 248 image for crack, 296 for dent, 276 for gesso line and 298 for paint issue, designing and implementing an algorithm by using a deep learning algorithm which utilizes a combination of Faster R-CNN as a detector and MobileNetV3 as a backbone network for feature extraction to automatically detect defects and the performance evaluation of the developed defect detection algorithm achieved defect detection accuracy of 100%. The evaluation of defects classification achieves accuracy of 68.13%, precision of 62.49%, recall of 68.13% and F1-score of 64.46%. Overall, this research aims to improve the efficiency and accuracy of quality control processes in the wood industry and enhances product quality and customer satisfaction.