A new rail inspection method based on deep learning using laser cameras

Santur, Yunus; Karaköse, Mehmet; Akın, Erhan

doi:10.1109/idap.2017.8090245

Cited by 43 publications

(24 citation statements)

References 14 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…The multi-task learning setting allowed the knowledge learned in one task to be transferred to another (i.e., transfer learning), forming a mutually beneficial mechanism. Similar work used CNN for railway track inspection [57], road pavement crack detection [58], and concrete crack detection [59].…”

Section: ) Supervised Learningmentioning

confidence: 99%

A Review on Deep Learning Applications in Prognostics and Health Management

et al. 2019

View full text Add to dashboard Cite

Deep learning has attracted intense interest in Prognostics and Health Management (PHM), because of its enormous representing power, automated feature learning capability and best-in-class performance in solving complex problems. This paper surveys recent advancements in PHM methodologies using deep learning with the aim of identifying research gaps and suggesting further improvements. After a brief introduction to several deep learning models, we review and analyze applications of fault detection, diagnosis and prognosis using deep learning. The survey validates the universal applicability of deep learning to various types of input in PHM, including vibration, imagery, time-series and structured data. It also reveals that deep learning provides a one-fits-all framework for the primary PHM subfields: fault detection uses either reconstruction error or stacks a binary classifier on top of the network to detect anomalies; fault diagnosis typically adds a soft-max layer to perform multi-class classification; prognosis adds a continuous regression layer to predict remaining useful life. The general framework suggests the possibility of transfer learning across PHM applications. The survey reveals some common properties and identifies the research gaps in each PHM subfield. It concludes by summarizing some major challenges and potential opportunities in the domain.

show abstract

Section: ) Supervised Learningmentioning

confidence: 99%

A Review on Deep Learning Applications in Prognostics and Health Management

et al. 2019

View full text Add to dashboard Cite

show abstract

“…In recent years, researchers and experts in the field of land and air transport have become increasingly interested in the application of artificial intelligence techniques to solve certain decision assistance problems, such as diagnosis of transport equipment, management of maintenance operations, analysis of driver behavior, prediction of deterioration of transport infrastructure, planning and forecasting of traffic demand, control of traffic signals, control of air traffic, etc. ; For example, machine learning has been used for rail maintenance forecasting [16], (fuzzy knowledgebased) expert systems for rail traffic control [17], deep learning for detection of lateral railroad defects [18], and neural networks for detection of defects on the surface of rails [19]. Meanwhile, big data analytics (BDA) has been used in particular to identify trends, discover relationships, implement predictive analysis, and give meaning to images, data flows, and various other types of information.…”

Section: Examples Of Applications Of Artificial Intelligence In Railwmentioning

confidence: 99%

Contribution of Artificial Intelligence to Risk Assessment of Railway Accidents

Hadj-Mabrouk

2019

Urban Rail Transit

View full text Add to dashboard Cite

In the design, development, and operation of a rail transport system, all the actors involved use one or more safety methods to identify hazardous situations, the causes of hazards, potential accidents, and the severity of the consequences that would result. The main objective is to justify and ensure that the design architecture of the transportation system is safe and presents no particular risk to users or the environment. As part of this process of certification, domain experts are responsible for reviewing the safety of the system, and are being brought in to imagine new scenarios of potential accidents to ensure the exhaustiveness of such safety studies. One of the difficulties in this process is to determine abnormal scenarios that could lead to a particular potential accident. This is the fundamental point that motivated the present work, whose objective is to develop tools to assist certification experts in their crucial task of analyzing and evaluating railway safety. However, the type of reasoning (inductive, deductive, by analogy, etc.) used by certification experts as well as the very nature of the knowledge manipulated in this certification process (symbolic, subjective, evolutionary, empirical, etc.) justify that conventional computer solutions cannot be adopted; the use of artificial intelligence (AI) methods and techniques helps to understand the problem of safety analysis and certification of high-risk systems such as guided rail transport systems. To help experts in this complex process of evaluating safety studies, we decided to use AI techniques and in particular machine learning to systematize, streamline, and strengthen conventional approaches used for safety analysis and certification.

show abstract

“…These systems are not affected by natural light because the cameras only acquire the laser wavelength. In many cases, this technique is applied in the manufacturing industry to obtain a complete reconstruction of an object using robotic arms [14] or by taking transversal sections of a piece while either the group of sensors or the piece are being moved [15]. In most cases, the aim of the systems is to measure specific features of the products, such as volume [16] or roughness [17].…”

Section: Introductionmentioning

confidence: 99%

Surface Defect System for Long Product Manufacturing Using Differential Topographic Images

Herrero

García

Usamentiaga

2020

Sensors

View full text Add to dashboard Cite

Current industrial products must meet quality requirements defined by international standards. Most commercial surface inspection systems give qualitative detections after a long, cumbersome and very expensive configuration process made by the seller company. In this paper, a new surface defect detection method is proposed based on 3D laser reconstruction. The method compares the long products, scan by scan, with their desired shape and produces differential topographic images of the surface at very high speeds. This work proposes a novel method where the values of the pixels in the images have a direct translation to real-world dimensions, which enables a detection based on the tolerances defined by international standards. These images are processed using computer vision techniques to detect defects and filter erroneous detections using both statistical distributions and a multilayer perceptron. Moreover, a systematic configuration procedure is proposed that is repeatable and can be performed by the manufacturer. The method has been tested using train track rails, which reports better results than two photometric systems including one commercial system, in both defect detection and erroneous detection rate. The method has been validated using a surface inspection rail pattern showing excellent performance.

show abstract

A new rail inspection method based on deep learning using laser cameras

Cited by 43 publications

References 14 publications

A Review on Deep Learning Applications in Prognostics and Health Management

A Review on Deep Learning Applications in Prognostics and Health Management

Contribution of Artificial Intelligence to Risk Assessment of Railway Accidents

Surface Defect System for Long Product Manufacturing Using Differential Topographic Images

Contact Info

Product

Resources

About