Subject: CGAL users discussion list
List archive
- From: <>
- To:
- Subject: [cgal-discuss] PCA 3D objects
- Date: Fri, 2 Mar 2012 16:17:55 +0100 (CET)
Hi everybody
I am using pca in order to solve problems related to pose, orientation in
matching
3D objects. For simetric objects, for instance a cube, if the faces of the
cube
are
perpendicular to the ortho axes the mean plane (by using
inear_least_squares_fitting_3) returns a plane
consilient with one of the main planes of reference. BUT if the cube is
rotated
(around Z axis, for instance counterclock pi/4), the returned plane is the X=0
(on Y axis). I can understand the this behavior because of the
linear_least_squares_fitting method and the fact "for PCA in CGAL it is the
default orthogonal projection." See
<http://cgal-discuss.949826.n4.nabble.com/linear-least-squares-23-td1013139.html#a1014773>.
My question is there is any undocumented function where is possible to achieve
the real principal component of the dataset, not related to the orthogonal
reference axes.
Best for all
ps. images can be added in order to help the understanding the problem
- [cgal-discuss] PCA 3D objects, robertolam, 03/02/2012
Archive powered by MHonArc 2.6.16.