Thread Subject: tracking/locating coordinates of an object with a specific colored pattern

Hi all, I am working in a research laboratory and need to use a pair of cameras to locate an object with a specific colored pattern in 3D space. The specific pattern I am using is wrapped around a cylindrical rod and is consisted of a yellow sticker with a strip of red and blue on either end and a spiral line through it.

Can someone give me some ideas of how to track this pattern using regular webcams? I am new to matlab and have done some research and reviewed some code about tracking a ball but seems like what I need to do is much more complicated because it require locating the position and twist of the rod to very great accuracy rather than just tracking it and putting a box around the object.

I am planning to use caltech's matlab camera calibration toolbox to find the internal and external parameters of the camera so the coordinate of each pixel can be related to actual 3d coordinates of the world. I am just not sure how to process an image and locate the 2D coordinates of the object in terms of the pixel value. How do you identify the pattern out of the image in the first place? How do you locate the image coordinates of the pattern I described above?
Is there a way of "erasing" everything else from the image and just leave the yellow sticker with red and blue ends?

Thanks for your help and pls give me suggestions!!

Hi,

I still haven't gotten a response from anyone. Please give me any input on this.
Anything would help me to think!

I found a paper using similar approach, their system perform edge detection and line scan on the image from both cameras and then use triangulation to locate the rod.

But that's not enough information to get me going because I am not sure how to locate the object in the image because the object may only be a small fraction of the entire image....


Please help!! THANKS!

Well how easy or hard it is really depends on what else is in the
image, doesn't it? A lot of background clutter, particularly if it
has the same colors and shapes as your desired target, will of course
make it more difficult. But you haven't supplied an image to us so
that's pretty much about all we can say, other than the useless vague
comments like you should use pattern recognition concepts, color
classification concepts, motion detection concepts, etc.

Overall your question reminds me of something I heard very very
recently, and that is this:
      "But that's not enough information to get me going ...."
Maybe you should think about WHY no one answered.

ImageAnalyst <imageanalyst@mailinator.com> wrote in message <fd4161a3-6208-406e-b79a-5b255f646b48@1g2000vbm.googlegroups.com>...
> Well how easy or hard it is really depends on what else is in the
> image, doesn't it? A lot of background clutter, particularly if it
> has the same colors and shapes as your desired target, will of course
> make it more difficult. But you haven't supplied an image to us so
> that's pretty much about all we can say, other than the useless vague
> comments like you should use pattern recognition concepts, color
> classification concepts, motion detection concepts, etc.
>
> Overall your question reminds me of something I heard very very
> recently, and that is this:
> "But that's not enough information to get me going ...."
> Maybe you should think about WHY no one answered.

To clarify my inquiry pls refer to the image here: http://img.photobucket.com/albums/v512/ninachenchen/Web%20Activities/marker.jpg

This is the pattern I would like to track, it's a yellow sticker with thin bands of red at two ends and a spiral pattern. The purpose of this rod is for surgery use. So the back ground would be mainly red (color of tissue and internal organs).

The existing tracking examples on internet mainly deal with identifying the object but does not compute the exact pixel coordinate.

For my case, the rod does not need to be moving, so this problem does not need to be at real time.

I hope I made my question clear enough.

Thanks :)