| Title |
Implementation of an AI Deep Learning-based Inspection System for Detecting Defective Vehicle Action Loaders |
| DOI |
https://doi.org/10.5370/KIEE.2023.72.12.1714 |
| Keywords |
Action loader; Deep learning; MobileNet; Defect inspection; Data augmentation |
| Abstract |
In this paper, we develop a detection system that combines machine vision and deep learning to obtain action loaders (systems consisting of actuators) and machine part images in real time with a camera. And the detection system informs the result that the learned machine can detect defects from the acquired image. As a deep learning engine, we propose a method of classifying grease application defects in production product images using MobileNet. Transfer learning and data augmentation techniques were used for insufficient defective image data. The proposed system accurately segments the defective part in the image, and uses the MobileNet model to detect the defect in the divided part. Experimental results show that the proposed method uses data augmentation techniques to demonstrate sufficiently high detection accuracy to be available in the field even in poor image data. |