On Wed, 5 Dec 2012, Greg Troxel wrote:
I have a large point-cloud. I would like to convert coordinate
systems, a 7-DOF rotate, translate, and scale.
My quick reaction is that gama isn't the right tool, since it's about
estimating coordinates given observations. If you already have an
internally-consistent point cloud, then you aren't estimating
coordinates.
I think your problem has two sub-parts:
estimating the rotation/translation/scale
applying it
Applying it is similar to datum transformation, except that datum
transformations are typically small angles.
It may be that part of the gama code is helpful.
Given only two control points, I would think you could
transform the control points real coordintes to ECEF XYZ (via proj4)
compute translation, scale
compute/choose an orientation, because you're down a control point
pretty easily, with the last two steps being done with a calculator
even.
Then, you could end up with rotate and translate matrices, and apply
them with octave.
If you had more control points, you'd be in a least-squares situation
(and better off data wise with a harder processing problem, really :-).
It may be that having framed the problem of multiple control points with
coordinates in two different systems, you can write code to use the
solver in gama to solve that different problem. But I'd expect the bulk
of gama to be about computing the partial derivatives of the types of
observations used in surveying.