Techniques for Automatic Inspection of Locomotive Pantographs
Pancam uses rapid two-dimensional image analysis techniques to measure wear and damage of pantographs. Pancam is implemented using the Halcon machine vision library combined with custom image processing software, a database, user interface and web interface.
Pancam inspects in-service locomotives. Individual coal trains typically contain 5 locomotives in sets of 2 or 3. When these trains are travelling at 80 km/h, the time between pantographs is as little as 360ms, which is too little time for an effective analysis. Pancam uses the time between locomotive sets and between trains to analyse the pantograph images, but rapid analysis is still important to ensure that all the pantograph images are properly analysed.
Pancam-captured image and annotated analysis showing notch damage to the carbon block.
Pancam-captured image and annotated analysis showing worn carbon blocks.
Pancam's inspection techniques are rapid and effective. Pancam uses the Halcon machine vision software library. Pancam uses Halcon's pattern matching techniques to identify the type of pantograph and locate it in the image. Side view images are analysed by a custom image processing algorithm that segments the pantograph from the background, followed by Halcon machine vision operators to identify notches and steps in the carbon profile that could damage the overhead wire. Excessively worn carbon blocks can also damage the overhead wiring, so worn regions of the carbon blocks are identified and reported.
Top view analysis verifying the pantograph horns.
To analyse top view images, Pancam again uses Halcon's pattern matching to identify and locate the pantograph. Pancam then recognises the horns and checks that they are not missing or bent out of shape because damaged or missing horns can also damage the overhead wire.
Pantograph inspection results are reported to a database with a web interface.Next: Website-based Reporting of Inspection Results.
Hamey Vision Systems Pty. Ltd. -- Innovative machine vision
A.B.N. 31 721 169 683