Automated visual inspection for magnetization process

Abstract

This paper presents the automated visual inspection method for classifying a magnet in the magnetization process of a spindle motor manufacturing. The original inspection process is performed by a trained human inspector. When an inspector handles the part manually, the risk of contamination in the process is usually high. The proposed automated visual inspection process will be integrated into the fully automated magnetizing machine that is currently developed in order to reduce the problem of contamination in the existing process. In this paper, the gray level image is converted to binary image using the thresholding level which is automatically calculated based on the principle of optimal thresholding histogram. Furthermore, two classification methods: the neural network approach and the template accumulator approach are proposed. These two methods were applied to this specific application and the results were compared.