Point-to-plane metric, point normals

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

Point-to-plane metric, point normals

Ivan Dryanovski
Hi everyone,

I'm interested in putting together a point-to-plane ICP variant using
PCL. I wanted to know if anyone has done something similar in PCL
before.

Related, is there any tutorial I can look at for extracting normals? I
found the following bit of code in the
"pairwise_incremental_registration.cpp" tutorial:

  pcl::NormalEstimation<PointT, PointNormalT> norm_est;
  norm_est.setSearchMethod (boost::make_shared<pcl::KdTreeANN<PointT> > ());
  norm_est.setKSearch (30);
  norm_est.setInputCloud (boost::make_shared<const PointCloud> (src));
  norm_est.compute (points_with_normals_src);

This looks pretty straightforward, but I'd be interested to learn more
about the details of the available search methods. Is digging in the
API the best way to learn that right now? Any tips are appreciated.

Thanks,
Ivan Dryanovski
_______________________________________________
[hidden email] / http://pointclouds.org
https://code.ros.org/mailman/listinfo/pcl-users
Reply | Threaded
Open this post in threaded view
|

Re: Point-to-plane metric, point normals

dirkholz
Hi Ivan,

I recently started splitting things up of what is currently in
pcl/registration (which is actually not that much), making it more
modular so to say so that we can have different methods for estimating
correspondences, rejecting correspondences, assigning weights to them
and, of course, having different error metrics and ways to optimize
them. This also means that I wanted to put point-to-point,
point-to-plane, plane-to-plane and alike in there. Feel free to join me
in these efforts. I'm currently prototyping a lot of these things with
Radu and wanted to put some first "results" here on the list for
discussion within the next days.

We might come up with some "task force" including all the interested
users and registration guys to get all these things together. So feel
free to join, just like all the pcl users/developers.

Best,
Dirk

On Wed, 2010-12-08 at 16:02 -0500, Ivan Dryanovski wrote:

> Hi everyone,
>
> I'm interested in putting together a point-to-plane ICP variant using
> PCL. I wanted to know if anyone has done something similar in PCL
> before.
>
> Related, is there any tutorial I can look at for extracting normals? I
> found the following bit of code in the
> "pairwise_incremental_registration.cpp" tutorial:
>
>   pcl::NormalEstimation<PointT, PointNormalT> norm_est;
>   norm_est.setSearchMethod (boost::make_shared<pcl::KdTreeANN<PointT> > ());
>   norm_est.setKSearch (30);
>   norm_est.setInputCloud (boost::make_shared<const PointCloud> (src));
>   norm_est.compute (points_with_normals_src);
>
> This looks pretty straightforward, but I'd be interested to learn more
> about the details of the available search methods. Is digging in the
> API the best way to learn that right now? Any tips are appreciated.
>
> Thanks,
> Ivan Dryanovski
> _______________________________________________
> [hidden email] / http://pointclouds.org
> https://code.ros.org/mailman/listinfo/pcl-users


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

Re: Point-to-plane metric, point normals

Ivan Dryanovski
Hi Dirk,

On Wed, Dec 8, 2010 at 4:35 PM, Dirk Holz <[hidden email]> wrote:

> Hi Ivan,
>
> I recently started splitting things up of what is currently in
> pcl/registration (which is actually not that much), making it more
> modular so to say so that we can have different methods for estimating
> correspondences, rejecting correspondences, assigning weights to them
> and, of course, having different error metrics and ways to optimize
> them. This also means that I wanted to put point-to-point,
> point-to-plane, plane-to-plane and alike in there. Feel free to join me
> in these efforts. I'm currently prototyping a lot of these things with
> Radu and wanted to put some first "results" here on the list for
> discussion within the next days.
>
> We might come up with some "task force" including all the interested
> users and registration guys to get all these things together. So feel
> free to join, just like all the pcl users/developers.

Sounds good, I'd like to stay in the loop if possible.

> Best,
> Dirk
>
> On Wed, 2010-12-08 at 16:02 -0500, Ivan Dryanovski wrote:
>> Hi everyone,
>>
>> I'm interested in putting together a point-to-plane ICP variant using
>> PCL. I wanted to know if anyone has done something similar in PCL
>> before.
>>
>> Related, is there any tutorial I can look at for extracting normals? I
>> found the following bit of code in the
>> "pairwise_incremental_registration.cpp" tutorial:
>>
>>   pcl::NormalEstimation<PointT, PointNormalT> norm_est;
>>   norm_est.setSearchMethod (boost::make_shared<pcl::KdTreeANN<PointT> > ());
>>   norm_est.setKSearch (30);
>>   norm_est.setInputCloud (boost::make_shared<const PointCloud> (src));
>>   norm_est.compute (points_with_normals_src);
>>
>> This looks pretty straightforward, but I'd be interested to learn more
>> about the details of the available search methods. Is digging in the
>> API the best way to learn that right now? Any tips are appreciated.
>>
>> Thanks,
>> Ivan Dryanovski
>> _______________________________________________
>> [hidden email] / http://pointclouds.org
>> https://code.ros.org/mailman/listinfo/pcl-users
>
>
> _______________________________________________
> [hidden email] / http://pointclouds.org
> https://code.ros.org/mailman/listinfo/pcl-users
>
_______________________________________________
[hidden email] / http://pointclouds.org
https://code.ros.org/mailman/listinfo/pcl-users