Re: a rigid transformation problem
- From: israel@xxxxxxxxxxx (Robert Israel)
- Date: 15 Mar 2006 22:14:15 GMT
In article <1142408468.680146.193720@xxxxxxxxxxxxxxxxxxxxxxxxxxxx>,
<jayzhuo@xxxxxxxxx> wrote:
given two point sets {pi} {qi}, determing the rotation matrix R and
translation matrix T, so that the error E is minimized:
E = square sum of d ( R*pi + T, qi)
Do you mean sum of the squares?
the distance is defined as follow:
d ( p, q ) = || p - q|| if || p - q|| < M
M otherwise
where M is a constance.
That could be rather nasty, as it makes the objective function
non-differentiable. Would a smooth cut-off such as
d(p, q) = M tanh(||p - q||/M) work for your purpose?
how to solve this problem?
Using a numeric solver.
You have your choice of parametrizing the rotation matrix, e.g.
in terms of Euler angles, or using constraints R' R = I and
det(R) > 0.
Robert Israel israel@xxxxxxxxxxx
Department of Mathematics http://www.math.ubc.ca/~israel
University of British Columbia Vancouver, BC, Canada
.
- Follow-Ups:
- Re: a rigid transformation problem
- From: Thomas Mautsch
- Re: a rigid transformation problem
- References:
- a rigid transformation problem
- From: jayzhuo
- a rigid transformation problem
- Prev by Date: Re: name for fx = x / (x + 1) ?
- Next by Date: Re: Calculus XOR Probability
- Previous by thread: Re: a rigid transformation problem
- Next by thread: Re: a rigid transformation problem
- Index(es):