| Title |
Development of a CNN Based Inspection System for Model Classification of Action Loader and Its Defect Detection |
| Authors |
전병주(Byoung Ju Jeon) ; 김동헌(Dong Hun Kim) |
| DOI |
https://doi.org/10.5370/KIEE.2024.73.12.2355 |
| Keywords |
Action Loader; Deep Learning; Classification; Defect Detection; Xception; Inception |
| Abstract |
In this paper, we propose an AI inspection system based on deep learning to analyze the product image of the action loader (linear actuator system) inside an automatic car door handle. The proposed system aims to classify the model of the action loader and detect its defects. In addition, the paper compares performance evaluation using CNN(Convolutionary Neural Networks)-based Xception and Inception models, respectively. Due to low defective image data, data augmentation and transfer learning techniques are applied to improve the performance. As an experimental result, Xception showed comparatively better results than Inception.
These experimental results were confirmed using Accuracy, Precision, Recall, and F1-Score based on the Confusion Matrix performance indicator. The proposed CNN-based inspection system presents practical applicability in the field of automobile part classification and defect detection. |