GPS position wrong merged

ODM
#1

Dear,

I’m new to the forum, and thank you in advance for your cooperation.

I have a dji mini2 that I use for

photogrammetric survey.

I’m always flying in the same zone, and my job is to represent and monitor how the terrain shape canches during some land works.

I have already made several flight in the area and then processed with WebODM, without having problems.

The last flight session gives me some issues during the processing phase.

This is the configuration of the mission:

  • to plan the mission I use Dronelink,
  • 5 GCP, measured with a DGPS(in Qgis the GCP is well located on the map),
  • 90m flight altitude with nadir pointed camera.

While processing the images I have some images that are merged into a position that doesn’t make sense:

  • -1/-3 meters on the MSL, while the altitude in my interest area is about 200 meters

Some notes on the elaboration:

  • if I esclude the GCPs from the processing, the surface of the model have no hole or wrong zones
  • the GCPs are all correct and well associated in the images

I attached some images about the error.

I’m not very interested in wheteher the “wrong” zone is processed correctly, the more i’m concerned that if the tie processing is to be completelly wrong, it is not considered.
There are some parameters to configure how to discard some outliers processed position?

I thank everyone and available to provide further information

2 Likes
#2

Hello,
are there something not clear in my post?
Could I upgrade some information?

thanks

#3

Giacomo,
I think i understand that your results have an odd outlier. This might indicate a incorrect location tag on one or more of you photos. Perhaps you can remove the photos and reprocess? . It also might possibly indicate a mis-marked GCP. Note that for the best results using GCP’s, the points should surround the area of interest. Let me know if this helps. We want you to have sucessful results! M

#4

Hello,
I had checked the GCP’s location several times and seemed to me to be right marked and located.
Removing the photos (10 on 300 photos dataset) and reprocessing the reduced dataset, gives the result that other images (that on the before run was well located) are wrong located, and we repeat the process one or two more times.

The GSP are quite close to the border (not to much to be sure to find in several photos).

If could Help we can share the dataset o more information
Screenshot from 2022-07-29 13-03-50

Thanks
Best

#5

Giacomo,
Mmm. Im not sure what the problem is. If you can share your photos and processing settings. I may be able to find the issue. As i reread your posts this part is very telling

To me it really points to a bad GCP location as the tie points process fine in the non- GCP run. M