CASTGUARD VISION: SMART DEFECT RECOGNITION IN MANUFACTURING INSPECTION SYSTEMS

Abstract

The rapid growth of industrial automation has increased the need for reliable and efficient quality inspection systems, particularly in manufacturing sectors such as casting production. Traditional inspection methods rely heavily on manual observation, making them time-consuming, inconsistent, and prone to errors caused by human fatigue and subjective judgment. With the increasing availability of image-based industrial data, machine learning techniques provide a promising solution for automating defect detection and improving inspection accuracy. This study focuses on the identification and classification of defects in casting product images. Conventional inspection systems often struggle to maintain consistency and scalability when processing large volumes of images. Existing approaches primarily depend on human expertise or rule-based image processing methods, which require handcrafted features and lack adaptability to varying industrial conditions. As a result, these systems often fail to detect subtle defects accurately and cannot efficiently handle large datasets. To address these limitations, a machine learning-based image classification system is proposed using Multinomial Naive Bayes (MNB), Decision Tree Classifier (DTC), and Random Forest Classifier (RFC). The system performs image preprocessing through resizing, normalization, and feature extraction by converting images into flattened feature vectors. These features are used to train and evaluate the models, with performance measured using accuracy, precision, recall, and F1-score. The proposed system enhances defect detection by providing automated, accurate, and consistent classification. The use of Random Forest, an ensemble learning technique, improves prediction performance while reducing overfitting. Additionally, a graphical user interface (GUI) enables easy user interaction, making the system suitable for practical industrial applications. Overall, the proposed approach offers a scalable, reliable, and efficient solution for casting defect detection, contributing to improved quality control and operational efficiency.