Rosanswers logo


I'm not sure if I should ask this here or in the PCL mailing list but I'll ask here anyways.

I am trying to use kdtree_flann to match narf descriptors. I am extracting the descriptors in a similar way to the NARF PCL tutorial. The problem is when I try perform the NN search between two pointclouds (pointcloud A added as the input and point cloud B used as the query), all the features in B match to the same feature in A. The distances reported by kdtree are all 0 (zero).

  • I noticed that narf36 has a {x,y,z,roll,pitch,yaw} value which are all 0 (zero) for my features, is this why all my distances are zero when kdtree performs its search?

The .descriptor values themselves are different for each feature in A and B. I would have expected that kdtree would use the Narf::getDescriptorDistance function or something similar to compute the distances.

  • Does anyone know what distance metric it uses when you provide the PointCloud< Narf36 > as the data type to kdtree?
  • And is there a way to change this (to maybe even a custom distance function)?

Thanks in advance,

  • Joe

Originally posted by robojoe on ROS Answers with karma: 11 on 2011-07-26

Post score: 1

Original comments

Comment by Asomerville on 2011-07-26:
Also, you might get a better response if you break up your post into bullet points, and direct questions. Many people tend to pass over long paragraphs.

Comment by fergs on 2011-07-26:
You'll probably get a faster/better response on the PCL mailing list.


1 Answer 1


Rosanswers logo

The comments above are correct; this is best asked on pcl-users.

Originally posted by Mac with karma: 4119 on 2011-09-02

This answer was ACCEPTED on the original site

Post score: 2


Your Answer

By clicking “Post Your Answer”, you agree to our terms of service and acknowledge you have read our privacy policy.