From: xgeorgiou@pathfinder.gr   
      
   "a.sabrin"  wrote in news:1194607138.724535.33210   
   @v2g2000hsf.googlegroups.com:   
      
   > Hello!   
   >   
   > I have doing project, and I have a problem.   
   > I am working in pattern classification project to classify circle ,   
   > rectangle and triangle, and I use some feature such as compactness,   
   > roundness, eccentricity and no. of point of convex hull , but it not   
   > give me needed accuracy, so I looking for another features.   
   > Can somebody point me to some feature I can extract  it?   
   >   
   >   
      
      
   If you're classifying standard shapes like these, 2 or 3 good shape features   
   are more than enough, for   
   example number of corvers and principal inertia axes ratio (max/min) for   
   symmetry. If you still have   
   problems with the accuracy, perhaps the classifier you're using is not robust   
   enough (try using simple   
   NN after some tests with MDC).   
      
      
      
   --   
   Harris   
      
   --- SoupGate-Win32 v1.05   
    * Origin: you cannot sedate... all the things you hate (1:229/2)