Title: A CNN approach for online metal can end rivet inspection

Authors: Maurício Edgar Stivanello; Juliano Emir Nunes Masson; Marcelo Ricardo Stemmer

Addresses: Department of Metal-Mechanics (DAMM), Federal Institute of Santa Catarina, Florianópolis, Santa Catarina, Brazil ' Federal University of Juiz De Fora, Juiz De Fora, Minas Gerais, Brazil ' Department of Automation and Systems (DAS), Federal University of Santa Catarina, Florianópolis, Santa Catarina, Brazil

Abstract: Can end rivet fracture is an important defect type that may arise during the manufacturing of metal cans used in the food industry. Thus, an inspection procedure must be performed to remove the defective can ends from the production line. Previous approaches have demonstrated the possibility of performing an automated inspection. However, these approaches faced limitations associated with description and classification as they employed classical techniques. In this paper, a new machine vision-based method for online can end rivet inspection is described. In the proposed method the rivets are localised by using blob analysis, while the description and classification are entrusted to a convolutional neural network. The experiments carried out using images acquired under real conditions of use demonstrate that the proposed approach outperforms the results obtained in previous works.

Keywords: automated inspection; convolutional neural network; fracture detection; machine vision; pull tab rivet.

DOI: 10.1504/IJCAT.2022.127821

International Journal of Computer Applications in Technology, 2022 Vol.69 No.3, pp.282 - 290

Received: 16 Jul 2021
Accepted: 17 Oct 2021
Published online: 19 Dec 2022 *

Full-text access for editors Full-text access for subscribers Purchase this article Comment on this article