Re: PCL 1.8.1 Comparing PrincipalRadiiRSD descriptors

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

Re: PCL 1.8.1 Comparing PrincipalRadiiRSD descriptors

Sérgio Agostinho
Can you explain why subtracting both descriptors and computing the norm
is not adequate?

I'm failing to see the issue but I assume you already thought of that
but found it not adequate.


_______________________________________________
[hidden email] / http://pointclouds.org
http://pointclouds.org/mailman/listinfo/pcl-users
Reply | Threaded
Open this post in threaded view
|

Re: PCL 1.8.1 Comparing PrincipalRadiiRSD descriptors

nenadpol
This post was updated on .
Hi Sergio,

I am strugling to understand RSD descriptors.

I have point cloud representing pipe-cylinder with radius ca. 615

As I understand, computing RSD should give min and max radius for each point
in point cloud.

Min radius should be around cylinder radius while max radius vary between
min radius  and infinity/planeRadius.

Values I get as result are not even close to this.

Min radius vary a lot depending on input feature search radius.

Can you give me some pointers how to compute RSD corectly?

Let say I compute RSD corectly, should I ignore max radius (since we are
talking about cylinder) if I want to compare two descriptors ?

thanks
Nenad



--
Sent from: http://www.pcl-users.org/
_______________________________________________
PCL-users@pointclouds.org / http://pointclouds.org
http://pointclouds.org/mailman/listinfo/pcl-users
Reply | Threaded
Open this post in threaded view
|

Re: PCL 1.8.1 Comparing PrincipalRadiiRSD descriptors

Sérgio Agostinho

On 12-12-2017 06:49, nenadpol wrote:

> Hi Sergio,
>
> I am strugling to understand RSD descriptors.
>
> I have point cloud representing pipe-cylinder with radius ca. 615
>
> As I undersand, computing RSD should give min and max radius for each point
> in point cloud.
>
> Min radius should be around cylinder radius while max radius vary between
> min radius  and infinity/planeRadius.
>
> Values I get as result are not even close to this.
>
> Min radius vary a lot depending on input feature search radius.
If you know what to expect and the results are off it might be time to
come up with a synthetic example in order to debug the issue.
> Can you give me some pointers how to compute RSD corectly?
Sadly no. I never used it and never had to look into the implementation.

Some material found online
https://ias.in.tum.de/_media/events/rgbd2011/marton.pdf
http://robotica.unileon.es/index.php/PCL/OpenNI_tutorial_4:_3D_object_recognition_(descriptors)#RSD
http://robotics.usc.edu/~potthast/06299166.pdf   //search the doc for RSD
http://citeseerx.ist.psu.edu/viewdoc/download?doi=10.1.1.645.5367&rep=rep1&type=pdf 
//Section 6.1

> Let say I compute RSD corectly, should I ignore max radius (since we are
> talking about cylinder) if I want to compare two descriptors ?
I'm not sure I would do that without understanding first why you're
getting those Inf/NaN values first.

Cheers

_______________________________________________
[hidden email] / http://pointclouds.org
http://pointclouds.org/mailman/listinfo/pcl-users