Techniques to clean your point cloud in Metashape

[Shovelhead]

Im curious as what techniques is the community using to clean data inside of Agisoft?

So far im using;

• Filter by confidence
• Gradual Selection - Reprojection error
• Gradual Selection - Projection accuracy
• Gradual Selection - Reconstruction Uncertainty

Does anyone remove some photos via estimate image quality and pull out any below X value?

Any tips to generally improve my point cloud processing results that can be done in program?

[tazzo]

What kind of noise do you have?

[Shovelhead]

This is more of a general question and not project specific.

[dpitman]

Does anyone remove some photos via estimate image quality and pull out any below X value?

This is viable, yes.  But my images have generally all met my threshold, so I have mostly stopped checking them. But I am very conservative with flight speed in order to not have motion blurring, which seems to be the biggest culprit for poor images in uas image sets,  if there is one.

[tazzo]

Hello,

You can try to reduce the image overlap to see if this help in your case. It's usually effective in improving texture quality, geometry can benefit in some conditions too (ex. if sub-optimal images are discarded, etc..)

There is a mask function you can use after creating a mesh, you can redo the processing with these masks after.

Ensure that your cameras are correctly calibrated and that pre-processing software are not changing the shape, and add camera groups if the focal distance change (ex. by varying focusing distance).

[Shovelhead]

Thanks for the replies, anyone else?

[tazzo]

You may also be interested in looking at some workflow report:

https://pubs.usgs.gov/of/2021/1039/ofr20211039.pdf

[Shovelhead]

You may also be interested in looking at some workflow report:

https://pubs.usgs.gov/of/2021/1039/ofr20211039.pdf

Thank you very much for that, tons of good suggestions here.