Visual-Based Battery Labelling Quality Checker System Using Convolutional Neural Network

Having an auto labeling machine in the company is much faster and later determined by the company, but there are times when using an auto labeling machine results in incompatibility with the installation of battery labels. Apart from that, there are often claims from consumers that the label is not placed in the right place because the installation is done automatically. Based on this problem, we developed a machine that can detect the quality of label placement on batteries using machine vision. This machine vision technology is combined with the Convolutional Neural Network method. The system can detect label placement errors on batteries with a standard level of accuracy. The system can detect and classify three categories of battery label conditions with the average precision results for each class for no label batteries, rejected batteries and ok batteries respectively being 97.8%, 100% and 100%. The mean average precision (mAP) value produced by the detection model was 99.4%.