Development of line-scan vision system for print quality inspection

Abstract

In recent years, the fast-moving consumer goods industry, aligning with Industry 4.0 practices, has been incorporating more modern smart technology into the manufacturing system. To enable a factory to become a “smart factory”, there is a need for the machines to be capable of exchanging information autonomously and control one another. One of the crucial aspects of a supply chain is the quality control process, where defective products are sieved out by hand traditionally. However, given the huge volume of products being shipped out every day, there is a need for a more efficient system. The aim of this project is to develop an automated solution where a computer is able to accurately detect defects on a sample packaging and pinpoint the location. The usage of a vision system along with machine learning elements such as Google Colab and Neurocle will be incorporated throughout this project. Future work includes increasing the number of datasets to increase the machine learning model’s accuracy and precision. Also, another goal is to train the model to locate and differentiate if there are various defects present on the packaging.