Point Cloud classification by color

[Patribus]

Hello,

would it be possible to classify the point cloud by colors? This would be an excellent function for PhotoScan, since in many cases, one can differentiate/separate  objects just by color.

I would imagine a color palette, where you can create 'color-zones', each one would correspond to a classification of the respective points.

I would be happy to get a feedback in respect to this possible feature.

Best regards
Patribus

[Bruno Andrieu]

Hello

this would be a very usefull feature indeed.
With 2D images classifying pixels by colour has plenty of scientific applications. So this would be a nice extension to 3D images.


So ++

Bruno

[Darko]

YES !! That's the feature that would be great to classify roofs, dense vegetation canopy etc. Doing this together with the regular existing filter we should be able to get the true DTM..

+1, sure !!
 

[SAV]

+1

[Patribus]

Hello, I'm glad that there are some supporters for this feature.

My question to the Agisoft Team: Can we expect something in this direction in a near future?

Xanks & best regards.

[Patribus]

Hello, me again.

Question to the experts of point clouds in this forum:

Is there a point cloud software able to classify points out of the cloud by color?

If yes, which one?

Thanks.

[frnx]

+1 from me as well!

Something like "paint zones corresponding to each label, then let PhotoScan classify the whole point cloud" could be rather easily implemented with SVMs or other classification algorithms.

[balyoa]

+1 for me

A classification not only by color but by color range.... because to filter vegettaion there many subcolors of greens

Thanks

[Pixel UAV]

+1

This would be very useful, as we spend a lot of our time removing vegetation from our surveys.  Could it be combined with the current classification system?  So that trees may be removed based on their color and relative height to its surroundings.

[Patribus]

Dear Alexey,

after a long time away from the forum, I'm back to stress the importance of the feature proposed in this post for my work (and probably many others).

Has there been some development with this?

Cheers

[Darko]

Hello,

Just one more idea about the classification filter. It would be nice to be able to classify the lowest points in the defined area (variable) and put them in the separate layer. I think that this filter would filter out all the trees and branches in the low denisty forest.

Best regards,
Darko

[tkwasnitschka]

I was going to post this request right now. Have a look at the Meshlab select by color feature and maybe enable a similar feathering for hue, lightness and saturation of point colors.
Many thanks
Tom

[Patribus]

Dear Alexey,

some news on this.

You stated once, that you were considering integration such a function in PS.

Cheers

[jtuhtan]

CloudCompare has a decent point cloud classification algorithm (CANUPO). Although the algorithm it uses includes multi-scale geometrical properties of the cloud, I have used it to remove vegetation. Here is a wiki describing the classification procedure:

http://www.cloudcompare.org/doc/wiki/index.php?title=CANUPO_%28plugin%29

You can use Cloud Compare to filter and process models and point clouds (sparse or dense) from PS as well:

http://www.danielgm.net/cc/

Jeff

[Patribus]

Hi Jeff,

thank you for your hint.
 
But as I understand it, once I export the point cloud from PS it won't be possible to re-import it again. And I need it within PS.

@Alexey: Can I hope for some development in this direction in a near future? Almost one year ago you stated, that it would not be so complex to integrate a point selecting filter by color in PS and that it would appear in one of the following releases. However there was no improvement in this subject, as far as I did not miss something.

Again, I'm developing a new forestry application using photogrammetry and PS and the issue I presented here (http://www.agisoft.com/forum/index.php?topic=2775.msg14711) is a problem. I would so much need a color-dependent point selection tool within PS in order to be able to effectively work with PS. 

Best regards