Identification of Rail Surface Defects Based on One-Shot Learning

Albert Ji; Yang Thee Quek; Eugene Wong; Wai Lok Woo

doi:10.1007/978-981-19-2259-6_73

Identification of Rail Surface Defects Based on One-Shot Learning

Albert Ji^*, Yang Thee Quek, Eugene Wong, Wai Lok Woo

^*Corresponding author for this work

Computer Information Sciences Department

Research output: Chapter in Book/Report/Conference proceeding › Chapter › peer-review

Abstract

Rail track surfaces could suffer from defects such as abrasion and deformation, thus to ensure rail system safety, the conditions of rail tracks must be monitored. With the advancement of deep learning and computer vision technologies, automatic detection and classification techniques are being tested to replace or complement manual patrolling for productivity improvement. However, classic neural network approaches require a large amount of data which could be time-consuming and limit the application of deep learning techniques. This paper proposes applying one-shot learning using a Siamese convolutional neural network to the identification of rail surface defects. The results show the reduced requirement of training speed and possess potentials for real-time applications.

Original language	English
Title of host publication	6th International Conference on Intelligent Transportation Engineering (ICITE 2021)
Editors	Zhenyuan Zhang
Place of Publication	Singapore
Publisher	Springer
Chapter	73
Pages	823-832
Number of pages	10
Volume	901
ISBN (Electronic)	9789811922596
ISBN (Print)	9789811922589
DOIs	https://doi.org/10.1007/978-981-19-2259-6_73
Publication status	Published - 1 Jun 2022
Event	6th International Conference on Intelligent Transportation Engineering : ICITE 2021 - Virtual, Beijing, China Duration: 29 Oct 2021 → 31 Oct 2021 http://www.icite.org/icite2021.html

Publication series

Name	Lecture Notes in Electrical Engineering
Publisher	Springer
Volume	901
ISSN (Print)	1876-1100
ISSN (Electronic)	1876-1119

Conference

Conference	6th International Conference on Intelligent Transportation Engineering
Country/Territory	China
City	Beijing
Period	29/10/21 → 31/10/21
Internet address	http://www.icite.org/icite2021.html

Keywords

Rail surface defect
Defect classification
One-shot learning
Siamese network

Access to Document

10.1007/978-981-19-2259-6_73

Cite this

@inbook{8204a0cb4a624ea8ba633ad2ca84363b,

title = "Identification of Rail Surface Defects Based on One-Shot Learning",

abstract = "Rail track surfaces could suffer from defects such as abrasion and deformation, thus to ensure rail system safety, the conditions of rail tracks must be monitored. With the advancement of deep learning and computer vision technologies, automatic detection and classification techniques are being tested to replace or complement manual patrolling for productivity improvement. However, classic neural network approaches require a large amount of data which could be time-consuming and limit the application of deep learning techniques. This paper proposes applying one-shot learning using a Siamese convolutional neural network to the identification of rail surface defects. The results show the reduced requirement of training speed and possess potentials for real-time applications.",

keywords = "Rail surface defect, Defect classification, One-shot learning, Siamese network",

author = "Albert Ji and Quek, {Yang Thee} and Eugene Wong and Woo, {Wai Lok}",

year = "2022",

month = jun,

day = "1",

doi = "10.1007/978-981-19-2259-6_73",

language = "English",

isbn = "9789811922589",

volume = "901",

series = "Lecture Notes in Electrical Engineering",

publisher = "Springer",

pages = "823--832",

editor = "Zhenyuan Zhang",

booktitle = "6th International Conference on Intelligent Transportation Engineering (ICITE 2021)",

address = "Germany",

note = "6th International Conference on Intelligent Transportation Engineering : ICITE 2021 ; Conference date: 29-10-2021 Through 31-10-2021",

url = "http://www.icite.org/icite2021.html",

}

Ji, A, Quek, YT, Wong, E & Woo, WL 2022, Identification of Rail Surface Defects Based on One-Shot Learning. in Z Zhang (ed.), 6th International Conference on Intelligent Transportation Engineering (ICITE 2021). vol. 901, Lecture Notes in Electrical Engineering, vol. 901, Springer, Singapore, pp. 823-832, 6th International Conference on Intelligent Transportation Engineering , Beijing, China, 29/10/21. https://doi.org/10.1007/978-981-19-2259-6_73

Identification of Rail Surface Defects Based on One-Shot Learning. / Ji, Albert; Quek, Yang Thee; Wong, Eugene et al.
6th International Conference on Intelligent Transportation Engineering (ICITE 2021). ed. / Zhenyuan Zhang. Vol. 901 Singapore: Springer, 2022. p. 823-832 (Lecture Notes in Electrical Engineering; Vol. 901).

Research output: Chapter in Book/Report/Conference proceeding › Chapter › peer-review

TY - CHAP

T1 - Identification of Rail Surface Defects Based on One-Shot Learning

AU - Ji, Albert

AU - Quek, Yang Thee

AU - Wong, Eugene

AU - Woo, Wai Lok

PY - 2022/6/1

Y1 - 2022/6/1

N2 - Rail track surfaces could suffer from defects such as abrasion and deformation, thus to ensure rail system safety, the conditions of rail tracks must be monitored. With the advancement of deep learning and computer vision technologies, automatic detection and classification techniques are being tested to replace or complement manual patrolling for productivity improvement. However, classic neural network approaches require a large amount of data which could be time-consuming and limit the application of deep learning techniques. This paper proposes applying one-shot learning using a Siamese convolutional neural network to the identification of rail surface defects. The results show the reduced requirement of training speed and possess potentials for real-time applications.

AB - Rail track surfaces could suffer from defects such as abrasion and deformation, thus to ensure rail system safety, the conditions of rail tracks must be monitored. With the advancement of deep learning and computer vision technologies, automatic detection and classification techniques are being tested to replace or complement manual patrolling for productivity improvement. However, classic neural network approaches require a large amount of data which could be time-consuming and limit the application of deep learning techniques. This paper proposes applying one-shot learning using a Siamese convolutional neural network to the identification of rail surface defects. The results show the reduced requirement of training speed and possess potentials for real-time applications.

KW - Rail surface defect

KW - Defect classification

KW - One-shot learning

KW - Siamese network

UR - http://www.scopus.com/inward/record.url?scp=85131920378&partnerID=8YFLogxK

U2 - 10.1007/978-981-19-2259-6_73

DO - 10.1007/978-981-19-2259-6_73

M3 - Chapter

SN - 9789811922589

VL - 901

T3 - Lecture Notes in Electrical Engineering

SP - 823

EP - 832

BT - 6th International Conference on Intelligent Transportation Engineering (ICITE 2021)

A2 - Zhang, Zhenyuan

PB - Springer

CY - Singapore

T2 - 6th International Conference on Intelligent Transportation Engineering

Y2 - 29 October 2021 through 31 October 2021

ER -

Identification of Rail Surface Defects Based on One-Shot Learning

Abstract

Publication series

Conference

Keywords

Access to Document

Other files and links

Fingerprint

Cite this