 | | | | | |
Subject:
|
Stereo Vision
|
Newsgroups:
|
lugnet.robotics
|
Date:
|
Thu, 10 Aug 2000 10:05:13 GMT
|
Original-From:
|
Laurentino Martins <LMARTINS@MARKTEST.nospamPT>
|
Viewed:
|
1605 times
|
| |
|
Hi
As you may know my current project has everything to do with image recognition.
I was reading one tread in this newsgroup and suddenly it struck me that many of the problems I currently have in finding the contours of the objects can be overcome it I have stereo vision!
Then I figured that the only way to do that with just one camera was if I can place a set of mirrors in the front to the camera and split vertically the image in two, each set of mirrors giving an image a few centimeters apart.
Then all I have to do is to compare the left and the right parts of the images and find the differences between them.
Any thoughts about that?
Laurentino Martins
[ mailto:lau@netcabo.pt ]
[ http://www.terravista.pt/Enseada/2808/ ]
| | | | | | | | | | | | | | | |
Subject:
|
Re: Stereo Vision
|
Newsgroups:
|
lugnet.robotics
|
Date:
|
Fri, 11 Aug 2000 10:04:42 GMT
|
Original-From:
|
Laurentino Martins <lmartins@%saynotospam%marktest.pt>
|
Viewed:
|
1540 times
|
| |
|
This approach has several problems.
One is that the left image will show you smaller objects than the right, because the light travels a greater distance.
The second is that precision is very important and the mirrors must be fixed otherwise will make the image tremble a lot.
If you want to subtract two almost identical images to see the edges like I want to, one pixel of unalignment is too much.
Sorry the english
At 00:38 11-08-2000 Friday, Doug Weathers wrote:
> You might be able to save yourself some trouble with mirrors if you mount
> the camera to one side instead of in the middle.
>
> . .
> \ . \ .
> \....\ .
> \ . \ .
> . .
> ##### <-- camera lens
Laurentino Martins
[ mailto:lau@netcabo.pt ]
[ http://www.terravista.pt/Enseada/2808/ ]
| | | | | | | | | | | | | | | | |
Check this out ;)
Recognizing Three-Dimensional Objects by Comparing Two-Dimensional Images.
Daniel P. Huttenlocher, Liana M. Lorigo. IEEE Conference on Computer Vision
and Pattern Recognition. June 1996, San Francisco, CA, USA
http://www.ai.mit.edu/people/liana/cvpr-2col.ps.gz
____________________
Marco C. aka McViper
| | | | | | |
