Segmentation API

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

Segmentation API

Gabriel Corvalan Cornejo
Hi,

I am playing a bit with the RegionGrowing segmentation method, and I am surprise to see that segmentation module does not share a common API.
Example:
  • pcl::EuclideanClusterExtraction -> void extract (std::vector <PointIndices> &clusters)
  • pcl::RegionGrowing -> std::vector < std::vector <int> > getSegments () const

Is there a reason to these different API ?

Another issue is related to a missing feature in the RegionGrowing class. As in EuclideanClusterExtraction, Min and Max cluster size parameter could be useful. In my tests, about 90% of extracted clusters contains too few elements, so I discard them.


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

Re: Segmentation API

Radu B Rusu
Administrator
Gabriel,

This is a discussion better suited for pcl-developers@.

I believe the code in RegionGrowing hasn't been reviewed yet, so its API should be changed before including it in a
release to look more like the other segmentation methods. Any contributions along these lines are welcome. I believe
Sergey is the original author of the code, so you can also create an issue on http://dev and assign it to him.

Thanks,
Radu.

On 06/20/2012 03:31 AM, Gabriel Corvalan wrote:

> Hi,
>
> I am playing a bit with the RegionGrowing segmentation method, and I am surprise to see that segmentation module does
> not share a common API.
> Example:
>
>   * pcl::EuclideanClusterExtraction -> void extract (std::vector <PointIndices> &clusters)
>   * pcl::RegionGrowing -> std::vector < std::vector <int> > getSegments () const
>
> Is there a reason to these different API ?
>
> Another issue is related to a missing feature in the RegionGrowing class. As in EuclideanClusterExtraction, Min and Max
> cluster size parameter could be useful. In my tests, about 90% of extracted clusters contains too few elements, so I
> discard them.
>
>
>
> _______________________________________________
> [hidden email] / http://pointclouds.org
> http://pointclouds.org/mailman/listinfo/pcl-users
_______________________________________________
[hidden email] / http://pointclouds.org
http://pointclouds.org/mailman/listinfo/pcl-users