A GPU-based vision system for real time detection of fastening elements in railway inspection

Ruvo, P. De; Distante, A.; Stella, Ettore; Marino, F.

doi:10.1109/icip.2009.5414438

Cited by 47 publications

(39 citation statements)

References 5 publications

Supporting

Mentioning

Contrasting

Unclassified

Order By: Relevance

“…In [8], [9], Marino et al describe their VISyR system, which detects hexagonal-headed bolts using two 3-layer neural networks (NN) running in parallel. Both networks take the 2level discrete wavelet transform (DWT) of a 24×100 pixel sliding window (their images use non-square pixels) as an input to generate a binary output indicating the presence of a fastener.…”

Section: A Railway Track Inspectionmentioning

confidence: 99%

“…Year Components Defects Features Decision methods Stella et al [9], [18], [19] 2002-09 Fasteners Missing DWT 3-layer NN Singh et al [20] 2006 Fasteners Missing Edge density Threshold Hsieh et al [21] 2007 Fasteners Broken DWT Threshold Gibert et al [10], [11] 2007-08 Joint Bars Cracks Edges SVM Babenko [12] 2008 training examples from each class. This is necessary for the machine to learn a model that can handle variations in image appearance that result from changes in illumination, scale, rotation, background clutter, and so on.…”

Section: Authorsmentioning

confidence: 99%

See 1 more Smart Citation

Deep Multitask Learning for Railway Track Inspection

Gibert

Patel

Chellappa

2017

IEEE Trans. Intell. Transport. Syst.

335

158

View full text Add to dashboard Cite

Railroad tracks need to be periodically inspected and monitored to ensure safe transportation. Automated track inspection using computer vision and pattern recognition methods have recently shown the potential to improve safety by allowing for more frequent inspections while reducing human errors. Achieving full automation is still very challenging due to the number of different possible failure modes as well as the broad range of image variations that can potentially trigger false alarms. Also, the number of defective components is very small, so not many training examples are available for the machine to learn a robust anomaly detector. In this paper, we show that detection performance can be improved by combining multiple detectors within a multi-task learning framework. We show that this approach results in better accuracy in detecting defects on railway ties and fasteners.

show abstract

Section: A Railway Track Inspectionmentioning

confidence: 99%

Section: Authorsmentioning

confidence: 99%

Deep Multitask Learning for Railway Track Inspection

Gibert

Patel

Chellappa

2017

IEEE Trans. Intell. Transport. Syst.

335

158

View full text Add to dashboard Cite

show abstract

“…They achieved a classification accuracy of 90%. de Ruvo et al (2009) presented a GPUbased vision system to recognize rail fastening elements. Their system can improve the performance of a quad-core CPU implementation by about 287%.…”

Section: Introductionmentioning

confidence: 99%

Real-time monitoring of brake shoe keys in freight cars

Zou

et al. 2015

Frontiers Inf Technol Electronic Eng

View full text Add to dashboard Cite

Condition monitoring ensures the safety of freight railroad operations. With the development of machine vision technology, visual inspection has become a principal means of condition monitoring. The brake shoe key (BSK) is an important component in the brake system, and its absence will lead to serious accidents. This paper presents a novel method for automated visual inspection of the BSK condition in freight cars. BSK images are first acquired by hardware devices. The subsequent inspection process is divided into three stages: first, the region-of-interest (ROI) is segmented from the source image by an improved spatial pyramid matching scheme based on multi-scale census transform (MSCT). To localize the BSK in the ROI, census transform (CT) on gradient images is developed in the second stage. Then gradient encoding histogram (GEH) features and linear support vector machines (SVMs) are used to generate a BSK localization classifier. In the last stage, a condition classifier is trained by SVM, but the features are extracted from gray images. Finally, the ROI, BSK localization, and condition classifiers are cascaded to realize a completely automated inspection system. Experimental results show that the system achieves a correct inspection rate of 99.2% and a speed of 5 frames/s, which represents a good real-time performance and high recognition accuracy.

show abstract

“…Similarly, Stellaet al [2] employed wavelet transform and principal component analysis to preprocess railway images and used the neural classifier to recognition the missing fasteners. To achieve real-time performance, Ruvo et al [3] applied the error back propagation algorithm to model the fasteners and automatic fastener detection. However, the main work of the researches above aim at searching for missing fasteners and they are difficult to detect the partly worn fastener.…”

Section: Introductionmentioning

confidence: 99%

Railway Fastener Defects Recognition Algorithm Based on Computer Vision

Li¹,

Li²

2015

Proceedings of the 2015 International Conference on Electronic Science and Automation Control

View full text Add to dashboard Cite

Abstract-Railway fastener detection is an important task in railway maintenance to ensure safety. However, the earlier detection methods based on computer vision have good performance on missing fasteners, but they have weaker ability to recognize the partially worn ones. In this paper, we exploit the axis-symmetrical structure to generate the first and second symmetry sample of original testing fastener image, and integrate the first and second image for improved representation-based fastener recognition. The underlying advantages of the scheme are as follows: first, the symmetry image can somewhat overcome the difficulty that the lack of training and testing samples. Second, the symmetry image is helpful for representation-based fastener recognition and we can obtain an accurate judgment of the original testing image by integrate the corresponding judgments of two symmetry image. The experiment results show that our proposed method can achieve a rather high precision.

show abstract

A GPU-based vision system for real time detection of fastening elements in railway inspection

Cited by 47 publications

References 5 publications

Deep Multitask Learning for Railway Track Inspection

Deep Multitask Learning for Railway Track Inspection

Real-time monitoring of brake shoe keys in freight cars

Railway Fastener Defects Recognition Algorithm Based on Computer Vision

Contact Info

Product

Resources

About