Compare PCL results with NIST data

classic Classic list List threaded Threaded
1 message Options
Reply | Threaded
Open this post in threaded view
|

Compare PCL results with NIST data

Saeid
This post was updated on .
I have a dataset published by  NIST
<https://www.nist.gov/pml/engineering-physics-division/dimensional-metrology/algorithm-testing
(reference datset) to compare the least square algorithms with their
reference dataset for verification purposes. PCL does  not have least square
but it has RANSAC which removes the outliers and then compute the least
square. To compare the results of PCL with NIST dataset, I set a high
threshold, so RANSAC would consider all data points as inliner which
bascially turns RANSAC to a least square algorithm. The results of this
comparision is pretty interesting. For some datset PCL reports exactly the
same results that I expect from NIST dataset, but for some other dataset (I
uploaded them here), the results are very different. Can anybody explain to me why
they are different ? Iuse pcl::SACMODEL_CIRCLE3D as my model and
pcl::SAC_RANSAC as my method.   cir3.ds
<http://www.pcl-users.org/file/t499024/cir3.ds>   cir3.fit
<http://www.pcl-users.org/file/t499024/cir3.fit
instructionsfornist-l2-reference-pairs.pdf
<http://www.pcl-users.org/file/t499024/instructionsfornist-l2-reference-pairs.pdf

I attach the reference dataset to this post. The file with .DS extension
includes X, Y, Z of the 400 points, and the file with .FIT includes the
result of fitting. The last number in the .FIT file is the diameter of the
least square fit circle. For this dataset that I uploaded, PCL computes the
diameter =   356.697
and NIST diameter = 361.80966768000886515
Can anybody explain to me why they are so different?
It would be great if someone can fit a circle to their dataset using PCL and let us know their results.
Thank you



--
Sent from: http://www.pcl-users.org/
_______________________________________________
PCL-users@pointclouds.org / http://pointclouds.org
http://pointclouds.org/mailman/listinfo/pcl-users