Forum

Author Topic: Bad stitching with good Sequoia dataset  (Read 4692 times)

andrey.d

  • Newbie
  • *
  • Posts: 26
    • View Profile
Bad stitching with good Sequoia dataset
« on: August 15, 2017, 12:01:07 AM »
Hi!

I have the dataset of 1576 RGB images of field by Sequoia camera. It has a very good overlap - more 9 images for most points. And I use CameraAlignment with  RollingShutter options.
But result is bad. What is the problem? 

Please see Agisoft Photoscan report for details
https://share.skyglyph.com/index.php/s/vBPCTQxP0eU8BxA

GrinGEO

  • Sr. Member
  • ****
  • Posts: 262
    • View Profile
Re: Bad stitching with good Sequoia dataset
« Reply #1 on: August 15, 2017, 10:12:01 AM »
The GPS Position of the images have very large offsets as you can see in the report. I tlooks like the stitiching errors are all there where the camera position have been calculated wrong. I would try an entire new alignment from fresh....

Try to go in high mode

Alexey Pasumansky

  • Agisoft Technical Support
  • Hero Member
  • *****
  • Posts: 15118
    • View Profile
Re: Bad stitching with good Sequoia dataset
« Reply #2 on: August 15, 2017, 03:02:31 PM »
Hello andrey.d,

I think that you should use Height Field option for the mesh reconstruction instead of Arbitrary to avoid severe artifacts by the edges of the area of interest.
Best regards,
Alexey Pasumansky,
Agisoft LLC

andrey.d

  • Newbie
  • *
  • Posts: 26
    • View Profile
Re: Bad stitching with good Sequoia dataset
« Reply #3 on: August 16, 2017, 01:34:34 AM »
Hello andrey.d,

I think that you should use Height Field option for the mesh reconstruction instead of Arbitrary to avoid severe artifacts by the edges of the area of interest.
I tried this too.

I think I found the reason. There is series of deformed photos due to rollingshutter (see attachment)
Should photoscan correct this problem by optimizing alighment with rollingshutter setting?

Alexey Pasumansky

  • Agisoft Technical Support
  • Hero Member
  • *****
  • Posts: 15118
    • View Profile
Re: Bad stitching with good Sequoia dataset
« Reply #4 on: August 16, 2017, 10:50:36 AM »
Hello andrey.d

I can suggest to use higher tie point limit in the Align Photos dialog (like 4,000 - 8,000) and try using High accuracy. Also make sure that in the Camera Calibration window all distortion coefficients prior to the camera alignment are zeros.

During optimization you can use Rolling Shutter compensation flag to correct the related issues (many users report that have problems aligning RGB Sequoia camera images due to the rolling shutter effect).

In addition prior to the optimization I recommend to check if there are any misaligned cameras (could be "stacked" to the same position) and reset alignment for them.
Best regards,
Alexey Pasumansky,
Agisoft LLC

GrinGEO

  • Sr. Member
  • ****
  • Posts: 262
    • View Profile
Re: Bad stitching with good Sequoia dataset
« Reply #5 on: August 16, 2017, 11:08:17 AM »
how can it be to have so much rolling shutter issue? never seen something like this

andrey.d

  • Newbie
  • *
  • Posts: 26
    • View Profile
Re: Bad stitching with good Sequoia dataset
« Reply #6 on: August 16, 2017, 11:33:53 AM »
how can it be to have so much rolling shutter issue? never seen something like this
Strong wind and no gimbal