A deep learning-based method for aluminium foil-surface defect recognition
by Hui Wang; Chunhua Gao; Yongfa Ling
International Journal of Information and Communication Technology (IJICT), Vol. 19, No. 3, 2021

Abstract: In order to effectively detect and classify various defects on the surface of aluminium foil products, including contaminants, coining, shine marks and scratches, etc., the method of convolution neural network (CNN) is used and the detection of surface defects of aluminium product is realised by machine learning. Firstly, aluminium foil images are collected by CCD camera, and edge detection is performed on these images to obtain a complete picture area. Then, the robust principal component analysis (RPCA) method was used to perform underlying low-rank and sparse decomposition on the data matrix to obtain the defect areas in these images; Finally, using TensorFlow platform to build a CNN model, loading aluminium foil images and training to get the CNN model parameters, according to this CNN network model, the surface defects of aluminium foil images can be detected and classified in real-time. Numerical experiments verified that the proposed algorithm has the following advantages such as high accuracy, favourable expansibility and so on. It can also be easily applied into surface defect detection for other objects.

Online publication date: Mon, 13-Sep-2021

The full text of this article is only available to individual subscribers or to users at subscribing institutions.

 
Existing subscribers:
Go to Inderscience Online Journals to access the Full Text of this article.

Pay per view:
If you are not a subscriber and you just want to read the full contents of this article, buy online access here.

Complimentary Subscribers, Editors or Members of the Editorial Board of the International Journal of Information and Communication Technology (IJICT):
Login with your Inderscience username and password:

Username:        Password:          Forgotten your password?

Want to subscribe?
A subscription gives you complete access to all articles in the current issue, as well as to all articles in the previous three years (where applicable). See our Orders page to subscribe.

If you still need assistance, please email subs@inderscience.com