The best point cloud to mesh algorithm

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

The best point cloud to mesh algorithm

SirM2X
I have a point cloud of a segmented object in PCL that I want to convert to a mesh.
I've seen and tried the greedy projection approach (http://www.pointclouds.org/documentation/tutorials/greedy_projection.php) along with smoothing methods described in (http://www.pointclouds.org/assets/files/presentations/ICCV2011-surface.pdf).

The smoothing works fine and the mesh is relatively smooth. However, there are a lot of holes in the model! Is there a better way of creating a mesh that I don't know of?

Here (http://www.pcl-users.org/Point-Cloud-small-hole-filling-td2869476.html) somebody mentioned a method in test_surface.cpp. Haven't tried it yet but really need to know if it's any better than what I have.

Any help is greatly appreciated :)
Reply | Threaded
Open this post in threaded view
|

Re: The best point cloud to mesh algorithm

VictorLamoine
Administrator
Hello,

You should test poisson reconstruction, it produces a "watertight" mesh, but approximates the data and does NOT conserve the vertices positions (it moves the points).

I'm not sure if it's implemented in PCL but it is in meshlab, give it a try:
https://vimeo.com/20933872

Bye
Reply | Threaded
Open this post in threaded view
|

Re: The best point cloud to mesh algorithm

SirM2X
Thank you for your response Victor!
I have tried poisson reconstruction but the resulting mesh looks nothing like what I expect. Also, there's that extra part (or "extension" plane as mentioned in http://www.pcl-users.org/poisson-surface-reconstruction-td4026975.html).

Is the dissimilar result because the algorithm is moving the points around?

And yes, it is available in PCL : )