Publication: Machine vision-based nondestructive inspection of wood surface cracks
Loading...
Date
2024-07
Authors
Nurul Syahirah binti Ismail
Journal Title
Journal ISSN
Volume Title
Publisher
Abstract
Wood inspection is a crucial process in industries such as construction, furniture manufacturing, and wood processing, where identifying defects like cracks, knots, and pinholes is vital to ensure the structural integrity and aesthetic appeal of products. Traditional manual inspection methods are labor-intensive, time consuming, and prone to human error, leading to inefficiencies and quality control issues. To address these limitations, recent advancements have introduced machine vision-based non-destructive inspection methods, utilizing high-resolution cameras and sophisticated computer vision algorithms for accurate defect detection. Despite these technological advancements, significant challenges persist, particularly in detecting small and complex defects due to limited and diverse datasets used for training deep learning models. The need for computational efficiency in practical industrial applications further complicates the deployment of these models. This research aims to enhance the accuracy, efficiency, and reliability of automated wood defect detection systems by leveraging the YOLO (You Only Look Once) algorithm, specifically YOLOv8, combined with data augmentation techniques. YOLOv8 is renowned for its speed and accuracy in real-time object detection, making it suitable for industrial applications requiring rapid and reliable defect detection. This study assesses the effectiveness of data augmentation techniques in improving the detection capabilities of YOLOv8 and evaluates its performance in terms of precision, recall, mean average precision (mAP), and F1-score. The research also includes a comparative analysis of YOLOv8 against traditional methods like SVM and older neural network models, highlighting its superior performance in diverse environments. Initial findings demonstrate that YOLOv8, enhanced with data augmentation, significantly improves the detection accuracy and generalization capability of the model, particularly for small and complex defects. The results indicate a substantial improvement in precision, recall, and mAP compared to conventional methods, validating the effectiveness of the proposed approach. These findings underscore the potential of advanced deep learning techniques in revolutionizing wood defect detection, leading to better quality control, reduced financial losses, and greater efficiency in the wood industry.