Forum

Author Topic: background suppression from single mask not working correctly  (Read 1150 times)

carpekd

  • Newbie
  • *
  • Posts: 7
    • View Profile
background suppression from single mask not working correctly
« on: November 25, 2018, 07:23:31 AM »
I am trying to recreate the method shown here (see banana example): http://www.agisoft.com/index.php?id=49

I can download the sample images of the banana and run it through Photoscan and it works perfectly.  I want to do the same with different items, specifically getting the top and bottom of a part by flipping it over.  However, when I try this on my own part, the bottom and side images don't align properly with the top images.  I can manually add markers to each photo and realign one-by-one, but that is extremely time consuming. 

Here is the set of images that I took: https://drive.google.com/file/d/1WA2ROfO5Rhjc_OhdV9udeyGBmR5XRk2E/view?usp=sharing

And here are the steps I took in Photoscan 1.4.4 Standard (and Pro Demo)
1 - load photos
2 - rectangular select all of the first photo with no object (only background) and apply mask
3 - align photos with "apply mask to tie points"

The top side images align nicely and create a scan with no background, i just need to figure out where I'm going wrong with getting the side and bottom images to align correctly.  I thought I had enough overlap for the alignment to work, but I don't know.  Hoping someone could point me in the right direction.

Alexey Pasumansky

  • Agisoft Technical Support
  • Hero Member
  • *****
  • Posts: 14813
    • View Profile
Re: background suppression from single mask not working correctly
« Reply #1 on: November 26, 2018, 03:58:12 PM »
Hello carpekd,

I think you should improve the quality of the input data. If you are anyway using tripod-mounted cameras, then it shouldn't be that difficult to lower the ISO and increase the focal depth (by using higher f/stop values) - it would allow to get more details from the object's surface, thus increasing the number of matching points detected.

Also I could suggest to occupy the image frame space more effectively, if distance between the object and camera can be reduced.
Best regards,
Alexey Pasumansky,
Agisoft LLC