Surface defect detection of metal castings based on Centernet

Yihan Li; Hongzhi Jiang; Huijie Zhao; Xudong Li; Qi Wang

doi:10.1117/12.2667511

27 January 2023 Surface defect detection of metal castings based on Centernet

Yihan Li, Hongzhi Jiang, Huijie Zhao, Xudong Li, Qi Wang

Proceedings Volume 12550, International Conference on Optical and Photonic Engineering (icOPEN 2022); 125501K (2023) https://doi.org/10.1117/12.2667511
Event: International Conference on Optical and Photonic Engineering (icOPEN 2022), 2022, ONLINE, China

Abstract

In the production process of metal castings, certain defects are easy to appear on the surface. Before the castings are put into use, it is necessary to detect whether there are defects on the surface. In this paper, the defect detection of metal casting surface is taken as the research object, and the Centernet deep learning model is used to recognize the defect of casting surface image. Centernet does not use Region Proposal Network (RPN) and Non-Maximum Suppression (NMS) in the training process. It has the advantages of simple model, good detection effect, fast running speed and easy integration and deployment. By predicting the category heat map, the length and width map of the detection frame and the center offset of the detection frame with the same image size as the input model, the defect target can be extracted. In this paper, the image dataset is expanded by cropping, rotating, changing brightness and contrast of the images. For the pit and scratch defects of metal castings, the Mean Average Precision (mAP) is higher than 0.9 for the metal casting defect image dataset, and the software integration of the model is completed.

Citation Download Citation

Yihan Li, Hongzhi Jiang, Huijie Zhao, Xudong Li, and Qi Wang "Surface defect detection of metal castings based on Centernet", Proc. SPIE 12550, International Conference on Optical and Photonic Engineering (icOPEN 2022), 125501K (27 January 2023); https://doi.org/10.1117/12.2667511

ACCESS THE FULL ARTICLE

INSTITUTIONAL
Select your institution to access the SPIE Digital Library.

SELECT YOUR INSTITUTION

PERSONAL
Sign in with your SPIE account to access your personal subscriptions or to use specific features such as save to my library, sign up for alerts, save searches, etc.

PERSONAL SIGN IN

No SPIE Account? Create one

PURCHASE THIS CONTENT

SUBSCRIBE TO DIGITAL LIBRARY

50 downloads per 1-year subscription

Members: $195

Non-members: $335 ADD TO CART

25 downloads per 1 - year subscription

Members: $145

Non-members: $250 ADD TO CART

PURCHASE SINGLE ARTICLE

Includes PDF, HTML & Video, when available