And what does the "search" bit mean? Do you want a cover of rotations?
If so, a reasonable start is to just to uniformly sample a 4D sphere
(we have code for that). The second step would be to even out the
points a bit. A good way to do that would be to repeatedly, compute
the convex hull of the points and then move each point to the centroid
of its neighbors. CGAL has code to compute the convex hull. At least
for 3D points, such an operation typically ends up with a nice even
cover after a few iterations. I can write the code if you want as it
would be pretty quick.
On Jun 2, 2009, at 12:09 PM, Keren Lasker wrote:
yes - i meant that it would return Rotation3Ds
On Jun 2, 2009, at 5:20 AM, Daniel Russel wrote:
I don't understand what you are looking for. It sounds like it
might be an even sampling of the set of rotations but your
prototype below only returns one rotation...
On Jun 2, 2009, at 4:10 AM, Keren Lasker wrote:
hi all,
Has anyone implemented a function that generates a set of grid
search orientations ?
We have sphere based and cylinder based rotation sampling, but it
seems that there are better ways (quaternions based) that maintain
a fixed "distance" between all pairs of orientations.