Development of biologically and geometrically inspired methods for automatic detection of rail surface defects

Pao Yue-kong Library Electronic Theses Database

Development of biologically and geometrically inspired methods for automatic detection of rail surface defects

 

Author: Meng, Ling Qiao
Title: Development of biologically and geometrically inspired methods for automatic detection of rail surface defects
Degree: M.Sc.
Year: 2014
Subject: Railroad tracks -- Inspection.
Hong Kong Polytechnic University -- Dissertations
Department: Faculty of Construction and Environment
Pages: xi, 97 pages : illustrations (some color) ; 30 cm
Language: English
InnoPac Record: http://library.polyu.edu.hk/record=b2759127
URI: http://theses.lib.polyu.edu.hk/handle/200/7575
Abstract: The reliability of rail tracks is essential for the safety of rail transit, which plays a significant role in society and the economy. Safety statistics from the U.S. Federal Railroad Administration (FRA, 2002) indicate that track failure-related train accidents resulted in 2,700 derailments and US$441 million in direct costs during the decade from 1992-2002. In Hong Kong incident related to rail cracks occurred on the MTR in January and February of 2011, causing severe service disruptions. Thousands of commuters were affected, as trains were delayed for several hours during the morning rush hour. In the past decade, China has developed a huge high-speed railway network in addition to its existing wide-spread rail system. To ensure better rail maintenance and avoid adverse train incidents, there is an urgent need to develop effective techniques for rail defect detection. In this thesis, a novel multi-feature LEGION algorithm firstly used to process the rail imagery. This algorithm includes an image segmentation step based on the gray value feature and a subsequent image segment grouping step for the extraction of the rail boundary on the basis of the orientation feature. The multi-feature LEGION algorithm help to quickly analyze the imagery and identify the rail boundary and other interesting regions on the image, in this thesis, the interesting region is boundary of railway and defects. Automatic rail boundary extraction plays an important role in the entire automatic process, and rail boundary used as an important reference information in the subsequent rail surface feature extraction. We also use mathematical morphology to extract initial active contour of rail boundary and some defects. Then an improved GAC model is applied to each interesting region within the rail boundary for the refined boundary extraction of these regions. This improved GAC model in this thesis, we choose to use Chan-Vese model, and we also add Power-law and Negative transformation in this model. After comparative analysis of results, we use Power-law Chan-Vese model to detect. All of these methods are described in greater detail in the following sections. An integrated biologically and geometrically inspired approach capable of detecting and recognizing rail source defects from rail images collected in a complex environment is implemented in this dissertation. The algorithms developed in this project will greatly assist rail health analysts in rail surface defect detection, and evaluation. The research and development results obtained in this approach will contribute to the safety of rail transportation and maintenance in Hong Kong and mainland China. The algorithms developed therein can also potentially be transferred to and used in a variety of other industry applications, for example, defect detection on road or tunnel surfaces. The results of this project will also contribute to the advancement of image understanding, scene analysis.

Files in this item

Files Size Format
b2759127x.pdf 20.50Mb PDF
Copyright Undertaking
As a bona fide Library user, I declare that:
  1. I will abide by the rules and legal ordinances governing copyright regarding the use of the Database.
  2. I will use the Database for the purpose of my research or private study only and not for circulation or further reproduction or any other purpose.
  3. I agree to indemnify and hold the University harmless from and against any loss, damage, cost, liability or expenses arising from copyright infringement or unauthorized usage.
By downloading any item(s) listed above, you acknowledge that you have read and understood the copyright undertaking as stated above, and agree to be bound by all of its terms.

     

Quick Search

Browse

More Information