Forum

Author Topic: Problems in processing thermal data (FLIR XT2) with 1.5.3 versus 1.5.0  (Read 2977 times)

3DWinter

  • Jr. Member
  • **
  • Posts: 82
    • View Profile
Hello,
I have been using Photoscan for years and then Metashape when it came out, I am no novice. But today I witnessed certain behaviors I cannot explain.
In one machine I had Metashape 1.5.0 I processed thermal data (.tif)  form an XT2, with aircraft flying two passes (out and back with the right overlap) and it worked well.
Today I got a colleague of mine to purchase a new Metashape license, she downloaded the 1.5.3, tried processing the same data and it had significant alignment problems. Eventually, I noticed that "Adaptive camera model fitting" was not selected in advance during the alignment process. I was hoping that would resolve it, it helped but she was not able to align her model like I could with my (1.5.0). To prove the license version was not the problem I upgraded to 1.5.3, and then I started to have similar problems of data, not aligning. Assuming that nothing has changed in the alignment algorithm between versions deep in the Metashape code, then my only explanation is that I could see that certain default parameters are different between version (this is just an idea).

On another computer, I have 1.5.1 but I am also having problems during the aligning step.

Mosaicing thermal data with only two passes is a challenge, but I did process that data with 1.5.0 from two different directories and the result was consistently excellent.
1. is there a way to (downgrade)  my 1.5.3 to the 1.5.0 (I wish I have not upgraded).
2. Are there any initial parameters that changed between these versions?
        I think I was able to eliminate that the number of tie points, or limit tie points was an issue.

3. the 1.5.3 also complains about  "zero resolution" - I had no problem processing with the 1.5.0. I am 100% confident that overlap both forward-track and sidetrack are right, and I have processed the data successfully with 1.5.0 so I am clueless what is going on.
thanks for your help

« Last Edit: July 12, 2019, 09:06:28 PM by 3DWinter »

3DWinter

  • Jr. Member
  • **
  • Posts: 82
    • View Profile
Agisoft, any comments on this?
How can I get Metashape 1.5.0?
thanks

Alexey Pasumansky

  • Agisoft Technical Support
  • Hero Member
  • *****
  • Posts: 13211
    • View Profile
Hello 3DWinter,

Can you provide the project files related to the same datasets processed in 1.5.0 and 1.5.3 where the difference can be observed, so that we could check it?

Best regards,
Alexey Pasumansky,
Agisoft LLC

hluong

  • Newbie
  • *
  • Posts: 4
    • View Profile
Hi all,

I also encounter similar problems with aligning datasets using the latest version of Metashape.

I successfully obtained good quality orthomosaics with version 1.5.3 build 8407, but then I had to upgrade to build 8469 because there were other bugs that causes some (random) crashes. Since this update, I am not able to generate decent orthomosaics (somehow in the alignment process, it get messed up, even with or without GPS data). I am now reconstructing older datasets for which I obtained good results in the past. First try already failed, but it could be by coincidence (I am re-running my script several times).

But in the meantime, if there is a possibility to downgrade I also want to try that option, because now I am also kind of stuck here.

Kind regards

Alexey Pasumansky

  • Agisoft Technical Support
  • Hero Member
  • *****
  • Posts: 13211
    • View Profile
Hello hluong,

Can you please share the sample image set and projects processed in versions 1.5.3 (build 8407 and build 8469) that we can use in order to reproduce and investigate the problem?
Best regards,
Alexey Pasumansky,
Agisoft LLC

3DWinter

  • Jr. Member
  • **
  • Posts: 82
    • View Profile
Re: Problems in processing thermal data (FLIR XT2) with 1.5.3 versus 1.5.0
« Reply #5 on: August 06, 2019, 03:49:03 AM »
I am in the field and I am delayed in responding.
Linked below is an example of the thermal tif file.
As mentioned above with 1.5.0 thermal mosaicing was straight forward and I tested that it was not a fluke over and over.  To prove to my colleague that the version of metashape was inert for not aligning the same thermal data on her computer I upgraded to 1.5.3 and since then I cannot process that thermal data. Unfrotently I did not save projects with different builds.
Can you please provide a link to the 1.5.0 version?

https://drive.google.com/drive/folders/1UoQ5BbVFQiY6J21B1ZC3LiDcTo_BKeCV?usp=sharing

stihl

  • Sr. Member
  • ****
  • Posts: 408
    • View Profile

3DWinter

  • Jr. Member
  • **
  • Posts: 82
    • View Profile
Re: Problems in processing thermal data (FLIR XT2) with 1.5.3 versus 1.5.0
« Reply #7 on: August 08, 2019, 01:48:44 AM »
Thanks for that. But because I already have 1.5.3, the installation aborts, stating a newer version already exists. Do I need to remove the 1.5.3 in order to install 1.5.0?
Also, I was hoping to have the 1.5.0 in a separate location so I can process the imagery with both versions and compare the results.
Thoughts?
thanks again!
« Last Edit: August 08, 2019, 02:38:35 AM by 3DWinter »

Alexey Pasumansky

  • Agisoft Technical Support
  • Hero Member
  • *****
  • Posts: 13211
    • View Profile
Re: Problems in processing thermal data (FLIR XT2) with 1.5.3 versus 1.5.0
« Reply #8 on: August 08, 2019, 10:42:27 AM »
Hello 3DWinter,

You need to uninstall the newer version first.
Then install the older one and move the installation directory to the custom location. After that install version 1.5.3 to the default location. After that you'll be able to run different versions on the same computer accessing the program executable from different folders.
Best regards,
Alexey Pasumansky,
Agisoft LLC

3DWinter

  • Jr. Member
  • **
  • Posts: 82
    • View Profile
Re: Problems in processing thermal data (FLIR XT2) with 1.5.3 versus 1.5.0
« Reply #9 on: October 29, 2019, 11:28:31 PM »
After a busy field season, I got time to get back to this issue. I have 1.5.0 and 1.5.5 both installed on the computer.
Seems with both versions the model stability might be weak, as I run alignment on medium  (thermal data) and it works fine and then the run it again I get the message:
                  Some photos fail to align. Please reset alignment for incorrectly aligned photos and rerun the photo alignment.
Seems that key points are around 300 ish per picture- I think that is a bit little no!? Is there a way to get a summary of key points per picture? With thermal low contras imagery, running alignment on high will enable higher tie-points?
Are there any tools via the GUI or the Python console to get key and tie point statistics? Perhaps there are a few troublesome pictures with low key points or tie points...

If alignment crashed, by just resetting and trying again the alignment works as if there was never a problem.

Thoughts
« Last Edit: October 29, 2019, 11:41:57 PM by 3DWinter »