Help! Sentera 6x multispectral data of Orchard (900 images total)

[chris_hatcher]

Hi all, please help!

I have taken images of an Orchard/Field with the Sentera 6x multispectral sensor. The drone, DJI Inspire 1v2, has the light sensor, GPS and I also used a colour calibration card (all come with the Sentera 6x) for the project.

TLDR: I am trying to get an orthomosaic using WebODM and the multispectral setting. It failed with exit code 1 (see report (.txt) and files in google drive link below).

I have made orthos before with webODM but this is my first multispectral endeavour. I can’t find my particular problem in the forums, or a clear answer to my issues. If I have missed something, please direct me to the correct page. Is there a guide for Sentera 6x? I noticed in the ODM site that this camera was used for the coding, so I’m sure it is something simple I am not doing. Details below:

1+2. Link to images and report: See below, I’m not sure why google drive added some ._jPG files to the RGB folder but I left them there in case. Report shows multispec identified and as far as I can tell it couldn’t find a file which is uploaded.

https://drive.google.com/drive/folders/10VpFzx0DzvatmalP2h39QfWJ9NEwWwZk?usp=sharing

3. Did you fine tune any parameters? On previous attempts, yes, but not this one
4. If applicable, what browser are you using? What operating system? How did you install the software? Google chrome, Windows 10, manually via the github instructions with docker, gitbash etc

Specifications of computer:
Intel Core i7-7700K CPU 4.2 GHz
64GB RAM

Many thanks in advance for your assistance,
Chris

[RikF]

I’ll spin up a node on the Elephant (slow, but a lot of memory!) and see what happens.

[RikF]

Hmmm. I’ve tried running subsets and I keep getting an error along these lines:

s2p[p.filename] = filename_map[filename_without_band]
KeyError: ‘IMG_0028_670.tif’

Now, there is no such file (it has the _30), so I’m not sure what is happening there. I’ve run with/without the calibration images and the rgb jpegs and I get the same result each time.

[Hubertus_Paetow]

Had the same issue with a set of sequoia images. Try renaming all images so that the filename contains information about shot number and band, e.g. CAL_001_GRE. Tried it with part of your dataset (GRN) , finished in 19:09 with good result.

[chris_hatcher]

Thank you for trying with a few different variations

[chris_hatcher]

Thanks Hubertus, I will give this a go over the next few days and see how I get on. I don’t suppose you know of a handy way to group-rename images? I suspect I can find a way using R. I’ll google if one doesn’t come to mind before Monday.

What a great community!

I will report back once I have given it another try.

[Saijin_Naib]

If you’re using Windows, ReNamer by den4b is beyond compare, IMO.
https://www.den4b.com/products/renamer

If you’re on Windows 10, there is also a PowerRename PowerToy that uses regex.

GitHub - microsoft/PowerToys: Windows system utilities to maximize productivity

[RikF]

That’s going into my toolbox for so many things!

[chris_hatcher]

Update: I managed to get 5 bands (excluding RGB) to run without the calibration images after changing the names of the files. I used a simple script in R:

bluenames ← list.files(“0-Blue-475-30/”, pattern = “tif”, full.names = T)
file.rename(from = bluenames, to=sub(pattern = “475_30”, replacement = “BLU”, bluenames))

Any additional advice for successful run with the calibration “CAL_” images included?

ODM doesn’t like the RGB files for some reason (an RGB sensor on the 6x). I am iteratively changing the suggested parameter to see if it helps but so far no joy. If I can’t figure it out I may post again (it may be the dataset).

Thanks for your help @Hubertus_Partow and RikF.

[Hubertus_Paetow]

That´s good news. At the moment there is no way to use the calibration images for radiometric calibration in webodm. Also i don´t think it is possible to process the TIF and JPG files in a single task, you have to run a seperate task for the jpgs. BTW it doesn´t make sense to have the RGB and monochrome images in one single file. I can try to run the RGB task if it helps.
Hubertus

[chris_hatcher]

Running the RGB images would be very helpful, please can you try? I will continue my attempts tomorrow but I wasn’t making much progress. It’s odd that the other bands processed but the RGB didn’t. Perhaps it’s because of the resolution? The other wavelengths are lower MP (2 or 4) than the RGB 20MP sensor.

Regarding the calibration images, that’s a shame. I haven’t found anything on how to utilise them–I will contact Sentera. If anyone would like info that I find I am happy to share.

[Hubertus_Paetow]

I tried to process the rgb-images with different settings without success. The reconstruction fails although there is enough sidelap and overlap. Perhaps there is a problem with the rolling shutter. Sorry, no idea what to do.

[chris_hatcher]

Thanks for trying. I will get some more data (probably Friday, weather permitting) and see if the same issue occurs with RGB and go from there.

[chris_hatcher]

Regarding the calibration files using the reflectance panel of the Sentera 6x sensor.
So this took an interesting turn. Sentera, after a few emails, replied with “an engineer has given you a link” with a github link. If ODM doesn’t currently use the CAL files, the link they sent may be of use for the ODM developers. I didn’t know whether to post it here or not, because it is a python script that I am not equipped to use/understand and I assume is outside the scope of this forum.

If appropriate, can someone raise this with one of the developers? Happy to share with them the scripts if it will be of use to them.

The calibration using a reflectance panel is useful for repeated measures e.g. comparisons over time, as it standardises the images.

[RikF]

AFAIK that should be to pre-process the images using the calibration files. You then feed the calibrated images into ODM.