Sign in to follow this  
johnstanp

Principal Component Analysis and OBB generation

Recommended Posts

Well, I am using the PCA to compute OBBs. I wanted to test my implementation and decided to do it that way: -first of all, test the correctness of the implementation for a cube that has a unit volume, centered at the origin. -secondly, apply a rotation than a translation to the cube and see if the OBB computed has an orientation equal to the rotation, and a center located at a position equal to the translation vector. When I write "an orientation equal to the rotation", I actually mean that the axes of the box should be equal to the columns of the rotation matrix, or their opposite. Am I right to assume that simply rotating a set of points has no effect on the eigenvectors of the covariance matrix, other than rotating them?

Share this post


Link to post
Share on other sites
if you use some ellipsoid instead of box it should work in a way you wrote.
for cube it might give you really ugly results.
if you build your covariance matrix just with 6 vertices you might end up with eigenvectors pointing in some weird directions (diagonals of the box or something)

Share this post


Link to post
Share on other sites
I now remember that using PCA gives quite bad fit with squares...I'll try using a set of points located on an ellipsoid. Thanks for the answer.

Share this post


Link to post
Share on other sites

Create an account or sign in to comment

You need to be a member in order to leave a comment

Create an account

Sign up for a new account in our community. It's easy!

Register a new account

Sign in

Already have an account? Sign in here.

Sign In Now

Sign in to follow this