Electronics > Electronics

learning machine vision

(1/1)

dmehling:
I am wanting to implement a method for fairly accurate measurements of various objects and I was thinking that perhaps machine vision techniques would be more accurate then using IR rangefinders or sonar.  There will be very few variables so the task of measuring dimensions of objects would be fairly routine for my robot.  For example, if I had my camera set up in a fixed location and if a small object appeared in its view, I could determine the width or height of the object based on how many pixels composed the image of the object.  What should I do to begin learning how to implement some of the more basic machine vision methods? I was thinking I would use a web cam to capture the image and then write a program for processing the image.

Steve Joblin:
There is a pretty good tutorial in the tutorials section on Society of Robots.  Also check out CMUCam and RoboRealm.

dmehling:
Yes I have read through the tutorial and found it to be helpful but it doesn't give many specifics about how exactly to implement those techniques.  I'm looking for information on how to do the actual programming and what the best language is.  I was thinking I might give Visual Basic a try since I have the free express version.  I could be persuaded to try something else if the price is reasonable (less than $40 would be great).

JesseWelling:
You should be able to find something  ;)
http://www.cs.cmu.edu/~cil/v-source.html

Have you done any programing before?

Admin:
I think you should read a few books on the subject first. (seriously, this isnt a beginner level project :P)

For example, how would your robot know the difference between that object and the background? How would you know if the object is really big but far away, or really small but close up?

And even if your robot correctly determines the size of the object, what would your robot do with that info?

As for languages, I would recommend either C, C++, or Java, for this . . .

Navigation

[0] Message Index