Detection of Paper Bag Quality Production by Image Processing using Deep Learning Method

Abstract :

Ensuring product quality and reliability is essential in the dynamic environment of industrial manufacturing. Detecting defects during production is critical to prevent the delivery of faulty products to customers. Traditional quality control methods, while effective in certain cases, often lack the efficiency, precision, and adaptability required for modern high-speed manufacturing. Manual inspection, in particular, is prone to errors and reduced accuracy. Recent advancements in Deep Learning (DL) and Computer Vision (CV) offer promising opportunities for automated defect detection, with the potential to transform quality control processes. This study focuses on implementing a VGG19-based Convolutional Neural Network (CNN) for automatic quality assessment of paper bags using image processing, replacing manual inspection methods. The proposed system was trained and tested on a dataset of 1,729 images, classified into “OK” and “NOT OK” categories based on defect presence. The model achieved an accuracy of 95.26%, significantly outperforming skilled human inspectors, whose accuracy typically ranges from 72 80%. These results demonstrate the effectiveness of DL and CV in enhancing manufacturing quality control by delivering higher accuracy and consistency than traditional manual inspection.