Semi-supervised  rail defect detection from imbalanced image data

Siamak Hajizadeh; Alfredo Nunez Vicencio; David Tax

Semi-supervised rail defect detection from imbalanced image data

Siamak Hajizadeh, Alfredo Nunez Vicencio, David Tax

Research output: Chapter in Book/Conference proceedings/Edited volume › Conference contribution › Scientific

55 Citations (Scopus)

263 Downloads (Pure)

Abstract

Rail defect detection by video cameras has recently gained much attention in both
academia and industry. Rail image data has two properties. It is highly imbalanced towards the non-defective class and it has a large number of unlabeled data samples available for semisupervised learning techniques. In this paper we investigate if positive defective candidates selected from the unlabeled data can help improve the balance between the two classes and gain performance on detecting a specic type of defects called Squats. We compare data sampling techniques as well and conclude that the semi-supervised techniques are a reasonable alternative for improving performance on applications such as rail track Squat detection from image data.

Original language	English
Title of host publication	Proceedings of the 14th IFAC Symposium on Control in Transportation Systems, CTS 2016, Istanbul, Turkey
Editors	Tankut Acarman
Pages	1-6
Number of pages	6
Publication status	Published - 2016
Event	14th IFAC Symposium on Control in Transportation Systems - ITU Faculty of Architecture, Istanbul, Turkey Duration: 18 May 2016 → 20 May 2016 http://www.cts2016.org/en/

Conference

Conference	14th IFAC Symposium on Control in Transportation Systems
Abbreviated title	CTS 2016
Country/Territory	Turkey
City	Istanbul
Period	18/05/16 → 20/05/16
Internet address	http://www.cts2016.org/en/

Keywords

imbalance data
semi-supervised learning
rail image data
rail defect detection

Access to Document

TurkeyCTS2016_S_Hajizadeh_FINALAccepted author manuscript, 1.33 MB

Cite this

@inproceedings{fe47bca81a7047c783a0b906ca7722fc,

title = "Semi-supervised rail defect detection from imbalanced image data",

abstract = "Rail defect detection by video cameras has recently gained much attention in bothacademia and industry. Rail image data has two properties. It is highly imbalanced towards the non-defective class and it has a large number of unlabeled data samples available for semisupervised learning techniques. In this paper we investigate if positive defective candidates selected from the unlabeled data can help improve the balance between the two classes and gain performance on detecting a specic type of defects called Squats. We compare data sampling techniques as well and conclude that the semi-supervised techniques are a reasonable alternative for improving performance on applications such as rail track Squat detection from image data.",

keywords = "imbalance data, semi-supervised learning, rail image data, rail defect detection",

author = "Siamak Hajizadeh and {Nunez Vicencio}, Alfredo and David Tax",

year = "2016",

language = "English",

pages = "1--6",

editor = "Tankut Acarman",

booktitle = "Proceedings of the 14th IFAC Symposium on Control in Transportation Systems, CTS 2016, Istanbul, Turkey",

note = "14th IFAC Symposium on Control in Transportation Systems, CTS 2016 ; Conference date: 18-05-2016 Through 20-05-2016",

url = "http://www.cts2016.org/en/",

}

Semi-supervised rail defect detection from imbalanced image data. / Hajizadeh, Siamak; Nunez Vicencio, Alfredo ; Tax, David.
Proceedings of the 14th IFAC Symposium on Control in Transportation Systems, CTS 2016, Istanbul, Turkey. ed. / Tankut Acarman. 2016. p. 1-6.

Research output: Chapter in Book/Conference proceedings/Edited volume › Conference contribution › Scientific

TY - GEN

T1 - Semi-supervised rail defect detection from imbalanced image data

AU - Hajizadeh, Siamak

AU - Nunez Vicencio, Alfredo

AU - Tax, David

PY - 2016

Y1 - 2016

N2 - Rail defect detection by video cameras has recently gained much attention in bothacademia and industry. Rail image data has two properties. It is highly imbalanced towards the non-defective class and it has a large number of unlabeled data samples available for semisupervised learning techniques. In this paper we investigate if positive defective candidates selected from the unlabeled data can help improve the balance between the two classes and gain performance on detecting a specic type of defects called Squats. We compare data sampling techniques as well and conclude that the semi-supervised techniques are a reasonable alternative for improving performance on applications such as rail track Squat detection from image data.

AB - Rail defect detection by video cameras has recently gained much attention in bothacademia and industry. Rail image data has two properties. It is highly imbalanced towards the non-defective class and it has a large number of unlabeled data samples available for semisupervised learning techniques. In this paper we investigate if positive defective candidates selected from the unlabeled data can help improve the balance between the two classes and gain performance on detecting a specic type of defects called Squats. We compare data sampling techniques as well and conclude that the semi-supervised techniques are a reasonable alternative for improving performance on applications such as rail track Squat detection from image data.

KW - imbalance data

KW - semi-supervised learning

KW - rail image data

KW - rail defect detection

UR - http://resolver.tudelft.nl/uuid:fe47bca8-1a70-47c7-83a0-b906ca7722fc

M3 - Conference contribution

SP - 1

EP - 6

BT - Proceedings of the 14th IFAC Symposium on Control in Transportation Systems, CTS 2016, Istanbul, Turkey

A2 - Acarman, Tankut

T2 - 14th IFAC Symposium on Control in Transportation Systems

Y2 - 18 May 2016 through 20 May 2016

ER -