Forum

Author Topic: Strange DEM result (new uav with new process)  (Read 2378 times)

johan.jvr13

  • Newbie
  • *
  • Posts: 10
    • View Profile
Strange DEM result (new uav with new process)
« on: June 26, 2017, 05:00:41 PM »
Hi all,

We unfortunately crashed our old drone and decided to go with a DJI Inspire 2 with X5R camera.
Since the photos from Inspire 2 have coordinates and roll pitch yaws, I included them in the generation of my DEM.
It was the first time following this process with camera coordinates so I used the tutorial from the website.
Attached is a snippet of the end result.
All along the flight lines there are these blocks that are higher than the surrounding area.  I'm not entirely sure why or how this happened.

I thing there might be a problem with the orientation of the photographs.

I'm running the same project but with the Generic photo alignment without roll pitch yaw parameters the way I did it with our old uav.  Will update if there is any change.

Looking forward to any response or idea on how this could've happened.

Alexey Pasumansky

  • Agisoft Technical Support
  • Hero Member
  • *****
  • Posts: 14813
    • View Profile
Re: Strange DEM result (new uav with new process)
« Reply #1 on: June 26, 2017, 05:08:31 PM »
Hello johan.jvr13,

There's nothing attached to your post, but I think it might be better to exclude the orientation procedure from the georeferencing procedure and use only camera locations data. Then rebuild DEM.
Best regards,
Alexey Pasumansky,
Agisoft LLC

johan.jvr13

  • Newbie
  • *
  • Posts: 10
    • View Profile
Re: Strange DEM result (new uav with new process)
« Reply #2 on: June 26, 2017, 05:31:40 PM »
Hi Alexey,

My bad...
The attachment should be here now.

I'm doing that now to see f it'll work.  Hoping for the best  :D

SAV

  • Hero Member
  • *****
  • Posts: 710
    • View Profile
Re: Strange DEM result (new uav with new process)
« Reply #3 on: June 27, 2017, 08:09:46 AM »
Hi johan.jvr13,

Have you checked your image overlap? I can see some 'radial striping' in your DEM, which often occur in models where the image overlap is insufficient.

These blocks that are higher might be areas where you don't have enough information from images to get a proper reconstruction. For example, if they were houses or trees and you were flying at a set elevation, the image overlap on these objects that are 'sticking out' of the ground would be even lower compared to the ground. Hence, PS would not be able to model them.

To sum it up, it looks like an image overlap issue to me. Take more pictures or fly at higher elevation (reduces your resolution of your subsequent photogrammetric model though).

Regards
SAV

johan.jvr13

  • Newbie
  • *
  • Posts: 10
    • View Profile
Re: Strange DEM result (new uav with new process)
« Reply #4 on: June 27, 2017, 09:26:31 AM »
Thanks SAV,

I started to wonder that myself.
From the report I can see that it's specifically at the side overlap where this occurs.
Usually where there are 3 to 6 overlapping frames.

I looked at reports from our older uav and there was never this striping pattern with the same overlap.

Alexey, the processing without the camera orientations produced the same result.

We are flying again tomorrow so I'll try and use more overlap to see if that will work.

Alexey Pasumansky

  • Agisoft Technical Support
  • Hero Member
  • *****
  • Posts: 14813
    • View Profile
Re: Strange DEM result (new uav with new process)
« Reply #5 on: June 27, 2017, 10:02:02 AM »
Hello johan.jvr13,

You can send the project file with just alignment results as PhotoScan Archive file (PSZ format) to support@agisoft.com, so that we can check, if there are any other issues apart from overlap.
Best regards,
Alexey Pasumansky,
Agisoft LLC

johan.jvr13

  • Newbie
  • *
  • Posts: 10
    • View Profile
Re: Strange DEM result (new uav with new process)
« Reply #6 on: June 27, 2017, 10:48:40 AM »
Hi Alexey,

Thanks, but I found the problem.
On the tutorial there is a suggestion to optimize the rolling shutter at the camera alignment optimization step. Since we are using DJI I used that.
When running the project without optimizing rolling shutter everything works out.

We are using theInspire 2 with X5S (not R as said before) and we use GS PRO as our flight mission planner.
GS Pro has the uav always pointing in the direction of the first flight line.  Thus for the next one the uav will fly backwards.
As a result all the photos point in the same direction instead of along the line of flight.
Is this why the rolling shutter optimization didn't work?