[ros-users] pcl normals

[Radu Bogdan Rusu]

Michael,

We're looking into this. We did some preliminary tests and we were sure that the loss of precision is not that big, so 
we were alternating between Xf and Xd for some time. Obviously, Xf fits into SSE nicely, while Xd requires extra 
computation time. Good to hear about the bug report, let's see what they say. We did go with Xd for Moving Least Squares 
as there the loss was larger.

Can you please send me the dataset in question so we can test it here as well? That would be great. I'll start a thread 
offline as this might take a couple of iterations to solve (the old precision vs speed problem).

Thanks,
Radu.

On 07/22/2010 12:30 PM, Michael Krainin wrote:
> After switching from point_cloud_mapping to pcl, I noticed that the
> normals I was getting for point clouds were of noticeably worse
> quality.
>
> I ended up doing a side-by-side comparison of the outputs of
> cloud_geometry::nearest::computePointNormal and
> pcl::NormalEstimation::computePointNormal. I found differences in
> resulting normals of as much as 30 degrees.
>
> After further debugging, I traced the problem down to the performance
> of Eigen::SelfAdjointEigenSolver when using Matrix3f rather than
> Matrix3d (recall that point_cloud_mapping used Matrix3d for covariance
> and pcl uses Matrix3f). I have filed a bug report with eigen
> (http://bitbucket.org/eigen/eigen/issue/149/selfadjointeigensolver-matrix3f-vs).
>
> For now, I would very strongly recommend changing covariance_matrix_
> in pcl/features/normal_3d.h to an Eigen::Matrix3d and adjusting
> functions that use it accordingly. I have done so in my local copy,
> and can confirm that it fixes the problem.
>
> -Michael Krainin
> _______________________________________________
> ros-users mailing list
> ros-users at code.ros.org
> https://code.ros.org/mailman/listinfo/ros-users

-- 
| Radu Bogdan Rusu | http://rbrusu.com/