Newsgroups: sci.image.processing
Path: cantaloupe.srv.cs.cmu.edu!das-news2.harvard.edu!news2.near.net!news.mathworks.com!usenet.eel.ufl.edu!news-feed-1.peachnet.edu!gatech!newsfeed.pitt.edu!uunet!ftpbox!news.acns.nwu.edu!news.eecs.nwu.edu!k9ape
From: k9ape@eecs.nwu.edu (Sheldon L. Epstein)
Subject: Re: rotation invariant
Message-ID: <D59ADv.4r5@eecs.nwu.edu>
Sender: usenet@eecs.nwu.edu
Organization: EECS Department, Northwestern University
References: <3jhdps$llg@infoserv.rug.ac.be> <3jheb9$ltd@infoserv.rug.ac.be>
Date: Sat, 11 Mar 1995 03:15:30 GMT
Lines: 33

In article <3jheb9$ltd@infoserv.rug.ac.be>,
Tim Devreese <tdevrees@eduserv.rug.ac.be> wrote:
>
> I am doing a project on pattern recignition. The aim is that our robot 
> can find the object that are on the table. The biggest problem is that it 
> should not matter how the object is orientated. It is quit simple to find 
> a transformation that takes out the translation, but only the rotation is 
> quit a problem. We have been working with moments (HU and others), but we 
> don`t get the best result (it is slow to begin with, and since there is 
> quite a lot of noice ... ) To avoid these problems we develloped some 
> special algorithms but still.
> The system should work with a large set of objects.
>
> Any other suggestions.
>
> Thanks : tim.devreese@rug.ac.be
> (University of Gent, Belgium)
>

Hello Tim,

You might want to check out The Affine Transform - which is described
in a few books on Imaging and on Projective Geometry.  Stemple &
Kneebone's book on Algebriac Projective Geometry (1957, Oxford Press)
is a place to start - if you can find it.  Most of the math in this
area was done about 90 years ago when the photographic camera and
photographic projection systems were becoming popular.

Good hunting.

		Shel Epstein, k9ape@eecs,nwu.edu
		Chief Engineer
		Epstein Associates
