DEVELOPMENT OF AN ANN-BASED DEFECT DETECTION SYSTEM FOR PROCESS QUALITY OPTIMIZATION IN BOTTLING INDUSTRY

The study focused on a novel method for detecting defects in bottle products using convolutional neural network (CNN). tool was used to capture and analyze during the packaging of beverage products. It involved an Internet Things (IOT) system that contains both client- server-side system. client-side is raspberry pi, which captures sample its camera sends it over server processing. processes then outputs result client side indication. Two models were developed. showed first CNN model detected two states bottle, classified as defect good state. second detects up five bottle. observed training process has prediction accuracy 90.165% 85% live testing. There positive outcome labels most parameters used. Thus, development detection algorithm, this can form basis developing integrated vision-based tracking bottles bottling industry. These results provide potential information guide industries need improve their automation terms product tracing, thus, improving productivity. Keywords: Quality, Product, Convolutional Neural Network, Bottle, Raspberry Pi DOI： https://doi.org/10.35741/issn.0258-2724.58.4.28