Home // IARIA Congress 2024, The 2024 IARIA Annual Congress on Frontiers in Science, Technology, Services, and Applications // View article

Surface Defect Detection System for AI Vision-Based Press Formed Products

Authors:
Dong Hyun Kim
Seung Ho Lee
Jong Deok Kim

Keywords: Surface Detection; Press Formed Product; Anomaly Detection.

Abstract:
The appearance of a product is the first thing consumers evaluate for defects, making surface inspection crucial. Among these exterior products, surface inspection of press-formed products is still done manually by visual inspection, prompting exploration of solutions for surface inspection automation through machine learning systems to adapt to various on-site changes. For machine learning-based surface defect detection models, there is often insufficient defect data for training, and a small amount of defect data makes it difficult to improve the learning performance. Particularly, as manufacturing processes stabilize, defect occurrences decrease, making it time-consuming to collect desired defect training data. This paper proposes a method for training models for defect detection by using only normal product data to train the defect detection model. It identifies defects on the product surface by generating defect data from normal data input, calculating the difference between normal data through restoration, and identifying defects on the product surface through connection and separation.

Pages: 168 to 173

Copyright: Copyright (c) IARIA, 2024

Publication date: June 30, 2024

Published in: conference

ISBN: 978-1-68558-180-0

Location: Porto, Portugal

Dates: from June 30, 2024 to July 4, 2024