Quality Control on Manufacturing Computer Keyboards Using Multilevel Deep Neural Networks

Keyboards are held to high standards and those with defects cannot be considered as quality products. On the manufacturing line, many computer keyboard defects can appear, each of which needs to be handled differently. As a result, these defects need to be separated into their own category. Currently, default detection is done through the analysis of camera images. However, this is a challenging task due to the similarity shared by many defects and dusts in the images. In this paper, we propose a novel algorithm using multilevel deep neural networks to first group the defects and then distinguish each defect from others. The results show that the proposed deep learning network architecture can automatically and accurately classify keyboard defects. In this paper, we achieve the classification accuracy of 91.89%. This deep learning technique can also be extended to classify defects of other products.