proj-imim:
Optimising control points
Ben Kreunen 2001-Jan-05 06:57:55
At 09:20 PM 01/04/2001 -0800, you wrote:
>John Blommers wrote:
>
> > Hughe's script results in an error because there are 28 variables to
> > beoptimize but only 10 control points. Mathematically speaking you
> > needat least 28 variables. My rule of thumb is to have at least 4
> > controlpoints per picture, so for 10 pictures you should have 40
> > control points.
>
>That raises a question for me. If each control point is two constraints
>(x and y), and
>you have one anchor image, then for the second image a single control
>point would
>allow you to optimize yaw and pitch, and a second control point would
>allow you
>to optimize roll and field of view. Is it really necessary to have as
>many control
>points as John is recommending?
It helps.
Once you've established good values for v,a,b,c and you don't vary these
then you can cut down on the number of control points you use. A single
control point isn't sufficient in this case because adjusting FOV alone
could align the points, without having corrected yaw, pitch or roll.
Where you select your control points from will also have a bearing on your
end result, especially if you are varying v.
<http://www.bigbenpublishing.com.au/360/docs/tutorial/advanced/step1.html>
shows a few examples of selecting control points and what effect they have
on the end result.
Cheers
Ben
===================================
http://www.bigbenpublishing.com.au/
===================================