Design Tool of Deep Convolutional Neural Network for Intelligent Visual Inspection

4th International Conference on Applied Materials and Manufacturing Technology

材料科学;机械制造

Nagata, Fusaomi^1 ; Tokuno, Kenta^1 ; Otsuka, Akimasa^1 ; Ikeda, Takeshi^1 ; Ochi, Hiroaki^1 ; Watanabe, Keigo^2 ; Habib, Maki K.^3

Graduate School of Science and Engineering, Sanyo-onoda City University, 1-1-1 Daigaku-Dori, Sanyo-Onoda

756-0884, Japan^1

Graduate School of Natural Science and Technology, Okayama University, 3-1-1 Tsushima-naka, Kita-ku, Okayama

700-8530, Japan^2

Mechanical Engineering Department, School of Sciences and Engineering, American University in Cairo, AUC Avenue P.O.Box 74, New Cairo

11835, Egypt^3

关键词: Convolutional neural network;    Deep convolutional neural networks;    Image generators;    Manufacturing process;    Original images;    Similar image;    Vision based inspection;    Visual inspection;   

Others  :  https://iopscience.iop.org/article/10.1088/1757-899X/423/1/012073/pdf
DOI  :  10.1088/1757-899X/423/1/012073

学科分类：材料科学（综合）

来源: IOP

【 摘 要 】

Recently, convolutional neural networks (CNNs) are used essentially to classify images as it helps to cluster them by similarity and perform recognition. In this paper, a design tool that helps to develop different deep CNNs (DCNNs) is presented. As an example, a DCNN is designed by using the developed tool to use it for vision based inspection to recognize undesirable defects such as crack, burr, protrusion and chipping which normally occur in the manufacturing process of resin molded articles. An image generator is implemented to efficiently produce many similar images for training. Similar images are easily generated by rotating, translating, scaling and transforming original images. The designed DCNN is trained by using the produced images and then tested through classification experiments. The usefulness of the design tool and the basic performance of the designed DCNN are introduced.