Design of a Fuzzy Inference Based Robot Vision for CNN Training Image Acquisition

Abstract

With the promotion of Industry 4.0, artificial intelligence has recently been attracting attention in the manufacturing industry. In particular, the image recognition such as Convolutional Neural Network (CNN) are using active inspection and testing processes. However, the learning of image recognition such as CNN is using a large amount of images as training data and it takes a lot of time and effort to image acquisition. In this paper, we propose a fuzzy inference based robot vision system for the automatic image acquisition of CNN training image from a camera mounted to the robot arm. In addition, the proposed system is considering fuzzy inference to suppress the vibration of the servo motor in the robot arm to speed up the image acquisition process.