A cyber-enabled visual inspection system for rail corrugation

https://doi.org/10.1016/j.future.2017.04.032Get rights and content

Highlights

  • A cyber–physical visual inspection system for rail corrugation is brought forward.

  • A new rail localization algorithm is proposed to overcome abnormal illumination.

  • A local frequency representation for corrugation images is designed for recognition.

Abstract

Rail inspection is one of the most important tasks to guarantee the safety of a railway transportation system, and it requires advanced information technologies (e.g. cyber–physical system and cyber–physical–social system) to build intelligent inspection systems. This paper presents a cyber-enabled visual inspection system for rail corrugation, which includes an on-board image acquisition subsystem and a corrugation identification subsystem. In the corrugation identification subsystem, a track image captured by the on-board image acquisition subsystem is first segmented by the rail locating algorithm based on weighted projection profile (briefly as RLWP). And then each column of the segmented rail image is represented by local frequency features and identified as corrugation line or not by a support vector machine (SVM). Lastly, the rail image is judged as corrugation by integrating the recognized corrugation lines. The experiment results show that RLWP is robust and accurate to localize rail region even for uneven or abominable illumination. Moreover, the precision and recall of the proposed corrugation detection system are 98.47% and 96.50%, respectively. They are 25% and 1% higher than those of traditional methods. At the same time, the detection speed is doubly faster than that of the traditional approach. c

Introduction

3C (computing, communication and controlling) technologies have been extensively applied in railway industry. A modern railway transportation system is controlled or managed by its cyber-enabled systems using advanced technologies, such as internet of things (IoT), cloud computing, big data, artificial intelligence and social networks. A modern railway transportation system is a complex cyber–physical system (CPS), which constitutes physical systems and their corresponding cyber systems  [1], [2], [3], [4]. These physical and cyber systems are tightly integrated at all scales and levels to provide a wide range of innovative services and applications. Cyber systems are increasingly embedded in all types of railway physical parts, and make the railway system be more safe, intelligent and energy-efficient.

Railway infrastructures (e.g. rail, bridge and tunnel) are the fundamental physical components of a railway system. High-speed railways and heavy-haul railways require higher quality for the maintenance of various railway infrastructures, so rail inspection becomes one of the most important tasks for the railway industry  [5], [6], [7], [8]. A railway inspection system can be regarded as a typical cyber-enabled system  [2], and it includes four main aspects, i.e. data collection, information processing, decision making and maintenance optimization. As a case study, this paper presents a visual inspection system for rail corrugation based on the framework of CPS.

Corrugation refers to a phenomenon of periodical and wave-shaped irregularity, and appears along the longitudinal surface of rail heads. Visual appearance of a typical corrugation is shown in Fig. 3. Corrugation is a typical surface defect for rail heads, and it can cause sharp increase of wheel–rail force  [9]. For example, the dynamic load in the section of a rail with heavy corrugation is 2 times more than its static load. This large force would shorten the service life of rails and wheels. In addition, corrugation often makes serious traffic noise, so the rails with heavy corrugation are also known as ‘screaming rail’. The noise and vibration produced by corrugation not only make passengers feel uncomfortable but also influence the lives of the residents living along railways. So corrugation detection has attracted more and more attention in recent years  [9], [10], [11].

Corrugation is traditionally inspected by manual sampling measurement with a special calliper. Obviously, this method is subjective and inefficient. Nowadays, various automatic detection methods have been brought forward, including the chord measurement method and inertial reference method  [10], [12]. These methods, which are mostly based on mechanical and physical mechanisms, are efficient to measure the physical characteristics of corrugation, such as wave length and depth. They, however, are not efficient enough for routine inspection. This paper puts forward an efficient visual inspection system for rail corrugation (VIRC) based on the principle of cyber–physical system. VIRC first localizes the exact rail position from an original track image. Then it extracts the frequency feature vector for each column of the rail image, and judges whether the column is corrugation line or not based on SVM. Lastly, VIRC makes a final decision about whether corrugation exists or not in the input image based on the judgement results of all corrugation lines. To summarize, the main contribution of this paper includes the following three aspects:

  • 1.

    We propose a new rail locating algorithm, named Rail localization based on Weighted projection Profile (RLWP), which can effectively overcome the effect of abnormal illumination and accurately locate the position of rails.

  • 2.

    We bring forward a local frequency representation for corrugation images. This representation achieves better classification performance than the traditional global Gabor features  [13]. Furthermore, its computational efficiency is higher than that of Gabor features.

  • 3.

    We design an actual cyber-enabled corrugation inspection system, which meets the demand of track inspection task and is actually tested in some railway networks in China.

Section snippets

Related work

Generally speaking, a modern railway transportation system shows dynamic, open, interactive and autonomic features, and it is a representative of cyber–physical–social system, as shown in Fig. 1(a). It comprises the three spaces, i.e., physical space, cyber space and social space. Physical space includes trains, railway infrastructure (such as rail, bridges and tunnels), other facilities and environmental objects. Cyber space comprises physical sensor networks, social networks, train control

Overview of VIRC

VIRC includes image acquisition subsystem and corrugation identification subsystem. The image acquisition subsystem is mainly composed of hardware, and its main function is to capture rail images in real-time. The corrugation identification subsystem analyzes the obtained rail images and judges whether an image contains corrugation or not.

Rail locating based on weighted projection profile

Rail locating is the first step of corrugation detection. It divides rail out of a captured track image to eliminate the influence of background region in the subsequent analysis. An accurate locating algorithm is the basis of CIS, since the subsequent procedures only deal with the segmented rail area.

Generally, the grey values of rail area are greater than those of background, so the method named Rail localization based on Projection Profile (RLPP) is proposed in our previous work  [20]. RLPP

Experimental results and analysis

The image sets in this experiment are collected in actual railway of China by the image acquisition subsystem of Section  3.1. This experiment builds the following two datasets:

Dataset 1: this data set is used for evaluation of rail locating, including 400 track images which are arbitrarily chosen from a large image set. The rail area of each image is marked manually and the labelled results act as ground truth for performance evaluation.

Dataset 2: this data set is applied for the evaluation of

Conclusion

This paper has put forward the cyber-enabled visual inspection system for rail corrugation, which uses the local frequency features of rail corrugation and SVM classifier. The experimental results show that the proposed RLWP overwhelms the traditional RLPP. Furthermore, the proposed corrugation identification method achieves notably better performance, compared with the baseline that applies global Gabor feature and SVM  [13]. In addition, the detection speed of the proposed method is double

Acknowledgements

This work is partly supported by the Fundamental Research Funds for the Central Universities (2014JBZ003, 2016JBZ006), Beijing Natural Science Foundation (No. J160004), and National Key Technology R&D Program (No. 2014BAK02B07).

Qingyong Li received the B.Sc. degree in computer science and technology from Wuhan University, Wuhan, China, in 2001 and the Ph.D. degree in computer science and technology from the Institute of Computing Technology, Chinese Academy of Sciences, Beijing, China, in 2006. He is currently a Professor with the Beijing Jiaotong University, Beijing. His research interests include computer vision and artificial intelligence.

References (27)

  • B. Chen et al.

    Security analysis of urban railway systems: The need for a cyber-physical perspective

  • A. Rajhans et al.

    Supporting heterogeneity in cyber-physical systems architectures

    IEEE Trans. Automat. Control

    (2014)
  • A. Sheth et al.

    Physical-cyber-social computing: An early 21st century approach

    IEEE Intell. Syst.

    (2013)
  • Cited by (0)

    Qingyong Li received the B.Sc. degree in computer science and technology from Wuhan University, Wuhan, China, in 2001 and the Ph.D. degree in computer science and technology from the Institute of Computing Technology, Chinese Academy of Sciences, Beijing, China, in 2006. He is currently a Professor with the Beijing Jiaotong University, Beijing. His research interests include computer vision and artificial intelligence.

    Zhiping Shi received the B.S. degree in engineering at Inner Mongolia University of Technology in Hohhot, China in 1995, the M.S. degree in application of computer science from Inner Mongolia University, China in 2002, and the Ph.D. degree in computer software and theory from Institute of Computing Technology Chinese Academy of Science in 2005. He is an associate professor at College of Information Engineering of Capital Normal University in China. His research interests include image understanding, machine learning and formal method.

    Huayan Zhang is a Master student of the school of computer science and technology, Beijing Jiaotong University, Beijing, China. His research interests include multimedia and computer vision.

    Yunqiang Tan is a Master student of the school of computer science and technology, Beijing Jiaotong University, Beijing, China. His research interests include multimedia and computer vision.

    Shengwei Ren received the B.Sc. degree in physics from Inner Mongolia University, Huhehaote, China, in 1990 and the Ph.D. degree in information science from the Beijing Jiaotong University, Beijing, China, in 2010. He is currently a Senior Engineer with the Infrastructure Inspection Research Institute, China Academy of Railway Sciences, Beijing.

    Peng Dai is currently a Senior Engineer with the Infrastructure Inspection Research Institute, China Academy of Railway Sciences, Beijing.

    Weiyi Li is currently an Engineer with the Infrastructure Inspection Research Institute, China Academy of Railway Sciences, Beijing.

    View full text