CGAL users discussion list

[Yong Xiao <>]

I seem to find a bug of "linear_least_squares_fitting_3".
I am interesting in fitting quality of plane and line least squares
estimating of CGAL, so I find the source file
"c:\CGAL-4.0\include\CGAL\linear_least_squares_fitting_points_3.h" and
"C:\CGAL-4.0\include\CGAL\PCA_util.h" and locate the functions.
In linear_least_squares_fitting_points_3.h, it is said
"returns a fitting quality *(1 - lambda_min/lambda_max)*:"
// fits a plane to a 3D point set
// returns a fitting quality *(1 - lambda_min/lambda_max)*:
//  1 is best (zero variance orthogonally to the fitting line)
//  0 is worst (isotropic case, returns a plane with default direction)
template < typename InputIterator, 
           typename K >
typename K::FT
linear_least_squares_fitting_3(InputIterator first,
                               InputIterator beyond, 
                               typename K::Plane_3& plane, // best fit plane
                               typename K::Point_3& c,     // centroid
                               const typename K::Point_3*, // used for
indirection
                               const K& k,                 // kernel
                                                 const 
CGAL::Dimension_tag<0>& tag).
And the fitting quality in fact is computed by function fitting_plane_3
which is defined in PCAutil.h.

But in PCAutil.h, in the function fitting_plane_3, the fitting quality is
computed by *(1.0-lambda_min/lambda_middle) *. The comments
linear_least_squares_fitting_3 do not match the real computation methods.

Are the comments wrong? or the codes in PCAutil.h wrong?

Does the comments of 

--
View this message in context: 
http://cgal-discuss.949826.n4.nabble.com/A-bug-or-not-Questions-about-linear-least-squares-fitting-3-tp4655324.html
Sent from the cgal-discuss mailing list archive at Nabble.com.