Forum

Author Topic: Image masking by point projection  (Read 2407 times)

GeoGecco

  • Newbie
  • *
  • Posts: 31
    • View Profile
Image masking by point projection
« on: July 07, 2015, 11:33:03 AM »
Hey guys,

I use Photoscan a lot for UAV-Photogrammetry and it's always taking ages to mask the images when taking a few thousand pictures.
My idea was to add a mask to all pictures by selecting pre-aligned sparse points.

Here the idea for an algorithm:

- Add images, align them (this should be accuate enough)
- View sparse cloud and select all points from the area or the object you want to reproduce.
- click button 'add mask by object'
- go through every picture and find corresponding projection points.
      - calculate hull of these projected points (needs to be a concave hull)
      - the computed hull is the polygon which is the base for an inverted mask
      - add inverted mask to image
- options: add factor to adjust the size of the polygon which is found in every image

afte that, re-calculate the new sparse from images and do your work.

I think this would help a lot of people and should not be very hard to programm. Just a pythin script for your wiki would be great.
What do you think about that?

All the best


James

  • Hero Member
  • *****
  • Posts: 748
    • View Profile
Re: Image masking by point projection
« Reply #1 on: July 07, 2015, 01:02:06 PM »
agree this would be quite useful, however you can already get a similar effect by building a mesh based on the sparse cloud and then using tools->import->import masks->from model

GeoGecco

  • Newbie
  • *
  • Posts: 31
    • View Profile
Re: Image masking by point projection
« Reply #2 on: July 07, 2015, 05:06:17 PM »
Hey James,
thats pretty much what I was looking for =D
Well but it would be great if Agisoft would implement a threshold because when using the mesh from a sparse cloud, the edges sometimes cut the actual object .
Thanks