Bug in pcl/kdtree/kdtree_ann.hpp

classic Classic list List threaded Threaded
2 messages Options
Reply | Threaded
Open this post in threaded view
|

Bug in pcl/kdtree/kdtree_ann.hpp

Nico Blodow
Hi all,

we have hit a bug, for which we have attached a patch, however we're not
sure if it breaks other cases.

Basically, if the kd-tree is built from extract_clusters using a input
cloud _with_ indices, the subset of the point cloud data gets copied to an
ANNPointArray cloud_, however the "radiusSearch (int index..." function
treats the passed index as a index into the input_ cloud, not into the
copied subset cloud_.

So in essence, we used to get crappy/no euclidean clusters when using
indices. Maybe the proposed change should be put into a if(use_indices_)
block?

Cheers,
 -- Nico
_______________________________________________
[hidden email] / http://pcl.ros.org
https://code.ros.org/mailman/listinfo/pcl-users

indices_patch_kd_radius_search.patch (942 bytes) Download Attachment
Reply | Threaded
Open this post in threaded view
|

Re: Bug in pcl/kdtree/kdtree_ann.hpp

Radu B. Rusu
Administrator
Fixed! (r32906 I believe)

Cheers,
Radu.


On 09/27/2010 09:22 AM, Nico Blodow wrote:

> Hi all,
>
> we have hit a bug, for which we have attached a patch, however we're not
> sure if it breaks other cases.
>
> Basically, if the kd-tree is built from extract_clusters using a input
> cloud _with_ indices, the subset of the point cloud data gets copied to an
> ANNPointArray cloud_, however the "radiusSearch (int index..." function
> treats the passed index as a index into the input_ cloud, not into the
> copied subset cloud_.
>
> So in essence, we used to get crappy/no euclidean clusters when using
> indices. Maybe the proposed change should be put into a if(use_indices_)
> block?
>
> Cheers,
>   -- Nico
>
>
>
> _______________________________________________
> [hidden email] / http://pcl.ros.org
> https://code.ros.org/mailman/listinfo/pcl-users
_______________________________________________
[hidden email] / http://pcl.ros.org
https://code.ros.org/mailman/listinfo/pcl-users