Multispectral orthos creation

I’m working on a multispectral data set but the camera we used is from a firm named Hiphen and each image consists of 6 different spectral ranges.
which are
450nm
530nm
570nm
675nm
730nm
850nm
which are different from the cameras which were listed in the docs.

I have uploaded and processed MS images and in the report, no ortho ODM report
i couldn’t actually see any Ortho image
Later, I used QGIS to open the file and I was able to see the processed image. But I noticed multiple overlays on the right side of the image and the outputs for the DSM and DEM were not at accurate.
I tried to process the contours using the MS ortho but the results were not accurate.
I’m attaching the MS image i have processed.!

Do you guys have suggestions for me to process a better ortho?
or else is the problem due to the insufficient data for processing better ortho?
I do think it might be a problem due to the waypoints but I’m not quite sure.
for reference, I’m attaching the fight path details!
I tried to attach links but I couldn’t so did this
[Flight-path-MS — ImgBB
MS-Overlays — ImgBB
no-ortho-ODM-report — ImgBB]

I hope I can get a few suggestions or corrections I can do process better ortho…!
if the problem is with the flight path or some other issues, please do let me know, so I can be prepared for the next flight.
Guys your work has been amazing and Thank you…!

1 Like

Welcome!

Sorry it didn’t work great for you.

Are you able to provide your data for this camera somewhere like dronedb.app, as well as a link to the product page for the exact camera?

I likely have to see about adding support for it in OpenSFM at least.

I was able to get this info
and is there a place or link where I can find more information about Multispectral image analysis that can be done using WebODM such that I can distinguish the layers based on spectral ranges?
Ty…!

Could you give a bit more explanation or detail for what you want to accomplish?

I’m trying to obtain the images based on the spectral band!
currently, I’m using QGIS - singleband pseudocolor to obtain the images of respective bands.

I want to know whether I can do it just by using WebODM and if yes how can I include that in the report?

Thank you!

I’m not understanding still.

Do you have an example or mockup?

I’m sorry :grimacing:

Not a problem,

I’m asking is there a feature in WebODM that is similar to https://docs.qgis.org/3.22/en/docs/user_manual/working_with_raster/raster_properties.html#singleband-pseudocolor. this feature.

1 Like

Thank you! I’m sorry haha.

So, a way to view the individual bands of the processed datasets in WebODM UI?

No, not currently.

Cool!
I will process the latest data once again this week and update you!

Thank you

1 Like

Hello, Saijin!

We have processed a few more Multispectral datasets using WebODM and the results weren’t satisfactory. I tried to use another Software to process the images and it was able to obtain a decent ortho.
During this process, I was able to notice that the problem for stitching orthos in ODM may have occurred due to one of the following reasons.

  1. Existing algorithms might have referred to the Tiepoints of the images randomly rather than in sorted format.
  2. As multispectral have multiple spectral images may be a lack of coordination and aligning images might be a reason.
  3. I’ve noticed a problem with Georeferencing too, during the processing I was able to notice that the Geo-coordinates for the Multispectral images weren’t as accurate as they were for the RGB Data. This might be the critical issue for not building the Ortho.

Apart from this, An additional feature such as aligning the Images on the forehand and splitting the process into multiple steps can also save the processing time. With this method, we will be able to identify where the problem is.

These are mere suggestions based on my observations during the processing, I hope this helps. I would be glad if you got any suggestions or iterations for me to follow during the processing.

Best regards

1 Like

Additionally, it would be a better option to manually/Semi automate the process of changing the reflectance parameters based on the reflectance on the day images were acquired rather than automating it. Most of the multispectral sensor providers provide us with reflectance boards for calibrating the reflectance data.
Semi automate: Use the images which contain the reflectance board and extract the parameters from the images and substitute them for calibration.

1 Like

Are you able to share your dataset, processing parameters, and the console log?

Do you want me to share the datasets and processing parameters I worked on?
and the console log I’m not sure if I can do that, because I was processing that in the institute and we have to delete the files or logs that we created after the usage.

1 Like

Yeah, if the log isn’t sharable, that’s fine. It just helps me see what you saw.

If you can share the data and the processing parameters, yes, that’s wonderful.

dronedb.app if you can.

sorry, I couldn’t reply soon enough…!
MS32
Ms40ODM
One is what I processed using ODM and another one is by using other sources.

Link for the log: Json MS log - 5e5eafd4
Hope this helps, If you require any additional data let me know…!
Thank you, Have a great day…!

1 Like

Can you try without --fast-orthophoto, and maybe increasing --feature-quality, --min-num-features, and --matcher-neighbors 25% or so.

Actually, I did try using without fast ortho the results weren’t that different but next time I will try using matching neighbors by 25%.
I will update you in the coming week

1 Like