Automated railroad track inspection
Articles, Blog

Automated railroad track inspection

August 15, 2019


We’re the computer vision lab and what we’re
looking at here is a system for automatically detecting defects on railroad tracks and the
project we were faced with was they’re asking us, “Can you find defects automatically
with a computer?” and the problem is that many times humans need to be able to, uh,
detect defects and this takes a long time, ah, to be on the track and look at it, and
it becomes a very tedious process and you have to have an expert do this. The problem is that over time the human eye
gets stressed and also you might not be able to pick up all these defects, which are very
small. So what we’re interested in doing is “Can a computer do this?” So right
now what we’re doing is we’re outfitting a computer with cameras and infrared sensors
and lasers and we’re saying, “Can we find an intelligent way to learn how to detect
these defects, defects such as cracks on the rail, burns on the rail, missing fasteners,”
and we’re interested in ways that a computer could learn just as a human learns to classify
these. So over the years we’ve developed computer
vision learning methods that could automatically detect some of these defects and this is saving
us both time and money because some of these systems are becoming much cheaper and you
might not have to train an expert every time you want to detect one of these defects. In essence, we are feeding the computer slowly
some information and it learns overtime. Computers are becoming cheaper everyday and they’re
becoming faster, ah, so the idea is that we could mount a camera, which doesn’t cost
much, onto this computer and it could do amazingly quick processing at speeds which humans couldn’t
even think of doing. So the idea here is that by mounting some of these sensors that don’t
cost that much, we could save ourselves a lot in the long time.

No Comments

Leave a Reply