Unsupervised surface defect detection using deep autoencoders and data augmentation

Surface level defect detection, such as detecting missing components, misalignments and physical damages, is an important step in any manufacturing process. In this paper, similarity matching techniques for manufacturing defect detection are discussed. We are proposing an algorithm which detects sur...

Full description

Saved in:
Bibliographic Details
Main Authors: Abdul Mujeeb, Dai, Wenting, Erdt, Marius, Sourin, Alexei
Other Authors: School of Computer Science and Engineering
Format: Conference or Workshop Item
Language:English
Published: 2020
Subjects:
Online Access:https://hdl.handle.net/10356/137972
Tags: Add Tag
No Tags, Be the first to tag this record!
Institution: Nanyang Technological University
Language: English
id sg-ntu-dr.10356-137972
record_format dspace
spelling sg-ntu-dr.10356-1379722020-04-20T13:38:13Z Unsupervised surface defect detection using deep autoencoders and data augmentation Abdul Mujeeb Dai, Wenting Erdt, Marius Sourin, Alexei School of Computer Science and Engineering School of Electrical and Electronic Engineering 2018 International Conference on Cyberworlds (CW) Fraunhofer Research Center Engineering::Computer science and engineering Defect Detection Automatic Optical Inspection Surface level defect detection, such as detecting missing components, misalignments and physical damages, is an important step in any manufacturing process. In this paper, similarity matching techniques for manufacturing defect detection are discussed. We are proposing an algorithm which detects surface level defects without relying on the availability of defect samples for training. Furthermore, we are also proposing a method which works when only one or a few reference images are available. It implements a deep autoencoder network and trains input reference image(s) along with various copies automatically generated by data augmentation. The trained network is then able to generate a descriptor-a unique signature of the reference image. After training, a test image of the same product is sent to the trained network to generate a test image descriptor. By matching the reference and test descriptors, a similarity score is generated which indicates if a defect is found. Our experiments show that this approach is more generic than traditional hand-engineered feature extraction methods and it can be applied to detect multiple type of defects. NRF (Natl Research Foundation, S’pore) Accepted version 2020-04-20T13:22:05Z 2020-04-20T13:22:05Z 2018 Conference Paper Abdul Mujeeb, Dai, W., Erdt, M., & Sourin, A. (2018). Unsupervised surface defect detection using deep autoencoders and data augmentation. Proceedings of the 2018 International Conference on Cyberworlds (CW), 391-398. doi:10.1109/cw.2018.00076 https://hdl.handle.net/10356/137972 10.1109/cw.2018.00076 391 398 en SMA-RP4 © 2018 IEEE. Personal use of this material is permitted. Permission from IEEE must be obtained for all other uses, in any current or future media, including reprinting/republishing this material for advertising or promotional purposes, creating new collective works, for resale or redistribution to servers or lists, or reuse of any copyrighted component of this work in other works. The published version is available at: https://doi.org/10.1109/CW.2018.00076 application/pdf
institution Nanyang Technological University
building NTU Library
country Singapore
collection DR-NTU
language English
topic Engineering::Computer science and engineering
Defect Detection
Automatic Optical Inspection
spellingShingle Engineering::Computer science and engineering
Defect Detection
Automatic Optical Inspection
Abdul Mujeeb
Dai, Wenting
Erdt, Marius
Sourin, Alexei
Unsupervised surface defect detection using deep autoencoders and data augmentation
description Surface level defect detection, such as detecting missing components, misalignments and physical damages, is an important step in any manufacturing process. In this paper, similarity matching techniques for manufacturing defect detection are discussed. We are proposing an algorithm which detects surface level defects without relying on the availability of defect samples for training. Furthermore, we are also proposing a method which works when only one or a few reference images are available. It implements a deep autoencoder network and trains input reference image(s) along with various copies automatically generated by data augmentation. The trained network is then able to generate a descriptor-a unique signature of the reference image. After training, a test image of the same product is sent to the trained network to generate a test image descriptor. By matching the reference and test descriptors, a similarity score is generated which indicates if a defect is found. Our experiments show that this approach is more generic than traditional hand-engineered feature extraction methods and it can be applied to detect multiple type of defects.
author2 School of Computer Science and Engineering
author_facet School of Computer Science and Engineering
Abdul Mujeeb
Dai, Wenting
Erdt, Marius
Sourin, Alexei
format Conference or Workshop Item
author Abdul Mujeeb
Dai, Wenting
Erdt, Marius
Sourin, Alexei
author_sort Abdul Mujeeb
title Unsupervised surface defect detection using deep autoencoders and data augmentation
title_short Unsupervised surface defect detection using deep autoencoders and data augmentation
title_full Unsupervised surface defect detection using deep autoencoders and data augmentation
title_fullStr Unsupervised surface defect detection using deep autoencoders and data augmentation
title_full_unstemmed Unsupervised surface defect detection using deep autoencoders and data augmentation
title_sort unsupervised surface defect detection using deep autoencoders and data augmentation
publishDate 2020
url https://hdl.handle.net/10356/137972
_version_ 1681056503992156160