Gradual selection image count

[PAPA2137]

I need help with python scripting.
How to perform gradual selection and removal of these points based on image count criterion?

EDIT
I found the solution

Code: [Select]

f = PhotoScan.PointCloud.Filter()
f.init(chunk, PhotoScan.PointCloud.Filter.ReconstructionUncertainty)
f.selectPoints(10.1)

[Yoann Courtois]

Really helpful piece of script ! Thank you !

(btw your script is dealing with reconstruction uncertainty filter, not image count  )

[Milderinne]

Hello everybody,
I would like to add a question to that subject.
I am not quite sure if I am able to express my idea/problem.

How is it possible to calculate the reonstruction uncertainty before using the gradual selection filter.
Considering these piece of code:

Code: [Select]

f = PhotoScan.PointCloud.Filter()
f.init(chunk, PhotoScan.PointCloud.Filter.ReconstructionUncertainty)
f.selectPoints(10.1)
 I have the feeling that 10.1 will deliver a really sparse point cloud in some cases. Therefore, I would like to change that number according to the range of reconstruction uncertainty depending on the specific project and sparse point cloud.

Anyone any idea? Any help appreciated.
All the best!

[Yoann Courtois]

Hi Milderine !

I would suggest you to read that post which is dealing well with gradual selection tools : http://www.agisoft.com/forum/index.php?topic=5536

Indeed, the reconstruction uncertainty, as well as other gradual selection tools, hasn't linked with specific project.
I mean a value of 10 is the same for every projects.
Nevertheless, you main indeed need different filtering strength depending on the project.
Then, you main need to define your own threshold depending on the kind as survey you handled.

As a start point, I mainly use those values :
- Reprojection error : 1 to 0.5 (easiest to hardest filtering)
- Reconstruction uncertainty : 50 to 25
- Image count : 1 (no filter) or 2
- Projection accuracy : 10 to 8

Regards

[Milderinne]

Thanks for your reply and the specific numbers!
Do you have the feeling that those numbers work for varying projects?
Best regards

[Yoann Courtois]

What kind of project are you mostly handled ?
What kind of camera / drone do you use ?

[Milderinne]

What do you mean with what kind of project?
When you mean objects of interest: I concentrate on outdoor projects, which means vegetation but also streets and houses and so on.
I use a Nikon D5500.
Regards

[Yoann Courtois]

I meant objects of interest, accuracy aimed and so on ("Traditional survey" or "very accurate (mm) model" for example).

I guess, from what you say, the values I gave to you are a good base. Then you can evaluate if it's too or not enough hard and adjust them.
But remember those filters are only a step in the middle of other things to get the best you can from your photogrammetric survey. You won't change your final result a lot by taking 50 or 60 as reconstruction uncertainty threshold.

Regards !

[Milderinne]

Ok, you meant what I thought.

I will just try it out, but I had the feeling 10 was to intense.
I also have other filters included already.

Thank you for your thoughts!
Best regards!

[Yoann Courtois]

Hi !

Projection accuracy filter aims to delete points that are matched between pictures at very different scale.
I mean one close picture and one far from the object. It reduce a lot the noise of your spare cloud.
Actually it wouldn't delete lots of points in a survey handled at regular distance from the object, but rather on surveys that present very irregular distances !

What kind of other filters do you also use ? Inside PhotoScan process ? I would be interested !

Regards

[Milderinne]

Hi!
I also use filters for image quality and reprojection error.
Regards