Printed label defect detection using twice gradient matching based on improved cosine similarity measure
作者：
Highlights：
• Propose an effective latent defect candidate extraction algorithm to eliminate artifacts caused by local deformation.
• Take advantages of twice gradient matching and masks.
• Present an improved cosine similarity measure motivated by the human visual system.
• Experimental results on several real-world industrial datasets demonstrate our superiority.
摘要
•Propose an effective latent defect candidate extraction algorithm to eliminate artifacts caused by local deformation.•Take advantages of twice gradient matching and masks.•Present an improved cosine similarity measure motivated by the human visual system.•Experimental results on several real-world industrial datasets demonstrate our superiority.
论文关键词：Printed label defect detection,Gradient matching,Artifact,Non-rigid local deformation,Cosine similarity
论文评审过程：Received 30 August 2021, Revised 11 January 2022, Accepted 25 April 2022, Available online 10 May 2022, Version of Record 20 May 2022.
论文官网地址：https://doi.org/10.1016/j.eswa.2022.117372
