Forum

Author Topic: Tie points are not well masked in vegetation and water regions  (Read 2080 times)

Ryuseiken

  • Newbie
  • *
  • Posts: 21
    • View Profile
Tie points are not well masked in vegetation and water regions
« on: November 01, 2019, 03:57:22 PM »
Hello.
I am working with UAV photos over an area partially covered with trees and water.
I'd like to mask them out in alighment because they sometimes degrades the camera parameter estimation.
The problem is that "Mask tie points" option doesn't work well in my case.

Specifically, I still get some tie points on regions with tall vegetation or water, even when I mask the regions on at least one photo and enable the "Mask tie points" option in alignment. Those tie points are not displayed on the masked images (indicating that they are not tied with the masked images), but it's clear that they fall inside the masked area if they are projected to the masked image.
I do not observe this phenomena when I apply masks to flat regions on the ground in the same image set.

Could someone tell me why this happens?

Alexey explains about  "Mask tie points" function as:
Quote
This feature will not create any new masks, it is just meant to avoid any tie points being created beyond the masked areas.
on this page:
https://www.agisoft.com/forum/index.php?topic=11021.msg49743#msg49743
but I don't understand well what "beyond" means here.

Another problem I've got is that the tie points in regions surrounding the masked regions are reduced (thinned) by the mask. This also happens in flat regions.

I am using Metashape version 1.5.5 build 9097.
Thank you for your helps.


James

  • Hero Member
  • *****
  • Posts: 748
    • View Profile
Re: Tie points are not well masked in vegetation and water regions
« Reply #1 on: November 01, 2019, 05:56:33 PM »
points in unmasked images are only 'masked' if they are matched with points in the masked image.

so if a key point is identified in an unmasked image, and matched in another unmasked image, but it is a point that wasn't detected in, or not matched in, the masked image(s) then it will become a tie point that doesn't get masked even if it falls in the volume described by a mask in the masked image(s).

this can happen as image features will have different appearance or may be occluded in images taken from different positions.

a solution might be to add more masks, or remove the tie points manually (or using gradual selection if applicable) from the sparse cloud after alignment.

Ryuseiken

  • Newbie
  • *
  • Posts: 21
    • View Profile
Re: Tie points are not well masked in vegetation and water regions
« Reply #2 on: November 12, 2019, 02:39:50 PM »
Hello, James.
Thank you very much for your clear and informative expalanation.

Excuse me for the delayed reply. I am preparing a test CG project to confirm that
Quote
points in unmasked images are only 'masked' if they are matched with points in the masked image.
I will report the result when ready.

As for the solutions, in my experience, the four criteria available in gradual selection are not effective in discriminating submerged (water) area.
The practical quick solution then seems to be deleting the submerged points after the alignment.
But in this case, I worry that such points may degrade the camera parameter estimation in the alignment stage and thereby induce inappropriate tie point filtering in the stage, which cannot be restoreed afterwards.
The best but time-consuming approach may be making water mask for each image using external machine-learning programs.

Ryuseiken

  • Newbie
  • *
  • Posts: 21
    • View Profile
Re: Tie points are not well masked in vegetation and water regions
« Reply #3 on: November 13, 2019, 03:02:19 PM »
I have attached the report about the CG test described above: a simple test on the “apply mask to tie points” feature in Metashape.

In addition, I have uploaded the relevant files (images and Metashape projects) here:
https://drive.google.com/file/d/13PHUnGQYcQtUmaAyPBLlW21AFQgagTTg/view?usp=sharing

I hope you and Agisoft support kindly check them out :).