Forum

Author Topic: Thermal Orthomosaic Workflow  (Read 6856 times)

Resourceful

  • Newbie
  • *
  • Posts: 36
    • View Profile
Thermal Orthomosaic Workflow
« on: May 14, 2020, 10:39:30 PM »
Good afternoon,

Does Metashape have the ability to create orthomosaics with R-JPEG or JPEG thermal?  I am attempting to create an orthomosaic with 640 x 512 XT2 data, with no luck. 

So, I am looking for alignment settings, and the workflow to attain an ortho.

Thank you!

Resourceful

  • Newbie
  • *
  • Posts: 36
    • View Profile
Re: Thermal Orthomosaic Workflow
« Reply #1 on: May 15, 2020, 06:39:12 AM »
Making progress!

The aspect ratio is incorrect though, the panels are correct width in this depiction, but the height is incorrect.  I feel like this is alignment settings, but trial and error is getting tedious.

Let me know what you think, thank you!

...with further attempts to correct the sizing, alignment has effectively come undone and further processing is for the most part failing.

Any insight into this would be amazing, thanks
« Last Edit: May 15, 2020, 07:51:02 AM by Resourceful »

Arie

  • Full Member
  • ***
  • Posts: 134
    • View Profile
Re: Thermal Orthomosaic Workflow
« Reply #2 on: May 15, 2020, 03:19:04 PM »
It's been awhile since I last worked with thermal images but I remember having to input some camera calibration values, such as sensor size, pixel size etc., before running the alignment. If I remember correctly, the pixel size wasn't square but varied between length and height.
Good luck!

Resourceful

  • Newbie
  • *
  • Posts: 36
    • View Profile
Re: Thermal Orthomosaic Workflow
« Reply #3 on: May 15, 2020, 05:16:04 PM »
Thank you Arie,  this is very helpful.

As a level 1 thermographer I should probably be able to answer this question, but any chance anyone would know where to find the info on pixel size - sensor size for the XT2?

Thanks again.

Arie

  • Full Member
  • ***
  • Posts: 134
    • View Profile
Re: Thermal Orthomosaic Workflow
« Reply #4 on: May 15, 2020, 05:38:40 PM »
You should be able to calculate it yourself: sensor width (mm)/ horizontal resolution (pixel) and sensor height (mm)/vertical resolution (pixel).
Or just check whether Flir Tools shows that in the EXIF data.

Arie

  • Full Member
  • ***
  • Posts: 134
    • View Profile
Re: Thermal Orthomosaic Workflow
« Reply #5 on: May 15, 2020, 05:47:06 PM »
I just remembered that I also had difficulties finding any information on the sensor size, which is needed for calculating the pixel size. But with the information on the field-of-view in combination with the focal length, you can calculate the sensor size. There are some online calculators for that.
Cheers!

Resourceful

  • Newbie
  • *
  • Posts: 36
    • View Profile
Re: Thermal Orthomosaic Workflow
« Reply #6 on: May 15, 2020, 07:31:45 PM »
Thanks Arie, I think I found the numbers I was looking for.

It looks like the sensor measures 10.88 mm x 8.7 mm, pixels are 0.017mm approximate.  The XT2 manual states 17 micrometers for pixel coverage, that makes sense.  Now I am back to trial and error on alignment, but I'm hoping when I hit the alignment I am looking for that the ortho will scale correctly.


Resourceful

  • Newbie
  • *
  • Posts: 36
    • View Profile
Re: Thermal Orthomosaic Workflow
« Reply #7 on: May 15, 2020, 08:02:47 PM »
Further update:

Still not looking right.  Anyone from Agisoft have any insights - workflows?  I've been using Raptor Maps for most of these jobs, but for the smaller sites like this one I would like to use the tools I have.  If Metashape isn't it, please let me know.


Alexey Pasumansky

  • Agisoft Technical Support
  • Hero Member
  • *****
  • Posts: 14813
    • View Profile
Re: Thermal Orthomosaic Workflow
« Reply #8 on: May 16, 2020, 04:13:23 PM »
Hello Resourceful,

If you can send the dataset and the project with the alignment results to support@agisoft.com, we could check, how to improve the results.
Best regards,
Alexey Pasumansky,
Agisoft LLC

dowallin

  • Jr. Member
  • **
  • Posts: 58
    • View Profile
Re: Thermal Orthomosaic Workflow
« Reply #9 on: May 17, 2020, 04:54:35 AM »
I am following this thread with great interest. I've been trying to mosaic XT2 imagery for some time with very limited success. The info in this thread has been intriguing. When I go to Tools-Camera Calibration, the pixel size that comes in by default is 0.0919118 by 0.11489. I've tried changing this to 0.017 by 0.017 as suggested, then aligning. I've run the alignment process several times. For each run I've start from scratch by starting a new project, load images new each time and each time I gone into the camera calibration window and set the pixel size to 0.017.

Different images get aligned on each run and my sparse point cloud is multilayered.

One other wrinkle that may or may not be significant. I went into FLIR tools and loaded one of the images. FLIR tools reads this as an image taken with a 19 mm focal length. Metashape camera calibration reads it as a 13 mm focal length. Looking at the lens on my camera, I've confirmed that it is a 13 mm lens.

Screenshots from each run are shown in the attached file. I'd be grateful for any suggestions.

Regards,
David
 

dowallin

  • Jr. Member
  • **
  • Posts: 58
    • View Profile
Re: Thermal Orthomosaic Workflow
« Reply #10 on: May 17, 2020, 07:20:28 AM »
And as a follow up, after I generate a thermal mosaic using RJPG images, how do I convert this to a temperature layer? I understand the issue of emissivity. In the terrestrial environment, there is considerable variation in emissivity among cover types. But my study area includes a river and water has a consistent emissivity. I'd like to be able to get water surface temperature.

How can I do this? Or should I be storing images as TIFFs rather than RJPG?

Regards,
David
 

Resourceful

  • Newbie
  • *
  • Posts: 36
    • View Profile
Re: Thermal Orthomosaic Workflow
« Reply #11 on: May 18, 2020, 12:15:00 AM »
Hello David,

I have sent the dataset to Alexey and team, here's hoping they get back to me.  I went through the same thing, with trying 0.017 by 0.017 pixel size, and the alignments came undone for sure.  Funny enough, I ran the dataset on a different computer without changing pixel size, and the point cloud came out dimensionally correct, but the ortho is still elongated after the process.  I also had no problem with the 19mm lens automation, that's the lens I'm running.

As far as getting water temperature goes, are you flying during the night or day?  I would be very particular about getting yoru environmentals, as well as TReflect when you put the photos into FLIR tools.  I always let my clients know "it depends" when it comes to determining absolute values.  The ITC sUAS thermography course certainly answered a lot of questions, if you haven't attended I would suggest it when the COVID stuff blows over.

Now I'm looking for in depth Metashape training, for visual - multispectral and thermal mapping.


Resourceful

  • Newbie
  • *
  • Posts: 36
    • View Profile
Re: Thermal Orthomosaic Workflow
« Reply #12 on: May 21, 2020, 12:50:28 AM »
Good afternoon,

I would like to close the loop with this, has anyone else had any luck with thermal processing?  Alexey and team, I did email a link to this data last week, and haven't heard anything since.

If I have time, I will try some re-processing to see if I missed anything.


dowallin

  • Jr. Member
  • **
  • Posts: 58
    • View Profile
Re: Thermal Orthomosaic Workflow
« Reply #13 on: May 21, 2020, 11:37:06 PM »
So I should NOT change the pixel size to 0.017 by 0.017? And I sure do not understand why I get a different result every time I run this. Computers are pretty darned good at doing the same thing over and over and getting the same answer.

I'm flying during the day. Not sure what you mean RE "environmentals and TReflect." I've never taken the thermography class but I'll consider doing so when things open up a bit. It seems to me that if you know the emissivity (and for clear water this is easy), it should be straightforward to get surface temperature.

Your thoughts? What am I missing?


Resourceful

  • Newbie
  • *
  • Posts: 36
    • View Profile
Re: Thermal Orthomosaic Workflow
« Reply #14 on: May 21, 2020, 11:58:45 PM »
At this point I am not sure what to do with the pixel size either.  I haven't received guidance from Agisoft at this point, so I will be off to try this with Pix4D at this rate.  I've just been trying to avoid purchasing another 5K software package.  At this point I'm not sure what you or I are missing, because the results are inconsistent. 

In a previous post, I had mentioned that without changing pixel size, I got a point cloud that measured pretty close (the panels are 1 meter by 2 meters) but the ortho came out with panels at 80 cm by 2 meters.

Changing pixel size gave me that parabolic dish of a rooftop, with a similarly distorted ortho.

Environmentals:  Temperature, humidity, wind.  TReflect = Reflected temperature, you get that from a diffused reflector set on a like angle to your subject.  In FLIR tools set distance 0, Emissivity 1 to get this from your target.  Remember the MFOV of your camera, I would not be trying to deliver absolute values without this info.  Then your absolute temperatures start to get close to actual.  Pretty tough to do with a map, which is more of a qualitative tool if I could get one together.