loading
This Thread  New Post   Forum Index 

 
Trying to work out Fiducial module
Anonymous from United Kingdom  [99 posts] 		14 year
Hi STeven,

I have been trying to reproduce the fiducial module so that I can explain it.   Your help file is super, and I looked at your references.
Here are my guesses so far:
First do thresholding and binarize the image.
Do a erode/dilate (or open) to clean up the image.
Then do a blob filter cutting on Min and Max size (as per the fiducial module) I also cut on Quadrilateral perimeter but this is just a guess?
This picks up the fiducial pretty well.  
Now I look at the Hu Moments and calculate the RMS between these and the hu moments of the of the training image.  I cut based on the sum of the RMS distances.  
The orientation of the image is pulled from the difference between the orientation moment of the image and the test image.
However, this doesn't fix the angle ambiguity problem.  

Am I close?  Totally on the wrong track?  

Thanks!!!!!!
profmason
Anonymous 14 year
You are close. The module uses slightly different statistics but the idea is the same. The last angle ambiguity should be one of four 90 degree orientation. Ie a square should have straight edges but you are not sure which edge is the X or Y axis. The module deals with this by simply comparing all four orientations of a fiducial to that in the DB ... give that at this point you are not working with much information this is not too taxing.

You may also need to use the Affine module to align the detected fiducial back to a square as apposed to a quadrilateral that may be rotated in X or Y.

STeven.

This forum thread has been closed due to inactivity (more than 4 months) or number of replies (more than 50 messages). Please start a New Post and enter a new forum thread with the appropriate title.

 New Post   Forum Index