DJI Phantom 4 multispectral in non ideal conditions

[bravemaster3]

Hello everyone,

I have read through all discussions and tried all kind of things, and finally thought of posting a question.
I have a bunch of DJI Phantom 4 multispectral (DJI P4M) images to process, and despite following the articles from Agisoft and other platform on how to best process them, I am having the problems pointed out by others, like in this discussion: https://www.agisoft.com/forum/index.php?topic=13109.0&fbclid=IwAR0qiGzMmBoKLfOmR5TrM9mpISCIV7j9t_HjocFkpuO_yn82ce4jpiIQDNI
Maybe it is not the same kind of problem, but here is the thing:
I know I did not fly under ideal conditions, and had some direct sunlight in some images on a cloudy day.
However, I believe there is a problem in the way Agisoft uses reflectance panel values for calibration of DJI P4M maybe?
I am saying this because the orthophoto looks good with Sun correction only, but I get the image in attachment when I use a 20% spectralon panel. Note, the choice of that panel was made after plotting Sun sensor corrected values of 5%, 20%, 50% and 99% with their real values, and noticing that there is a ~linear relationship (measured vs actual reflectance) which curves into more exponential between 20 and 50% and above.

Anyway, the reason why I am suspecting a problem in the way Agisoft does the integration of a reflectance panel is double: Apart from what I mentioned earlier that the results are good without panel, I also tried to export sun corrected images from Agisoft and do reflectance panel correction per chunk myself in python and then run the rest of the processing in Opendronemap, and the result was good (unfortunately it is not a viable option for me as it takes long time and needs a lot of resources as I have a lot of missions to process).

I wanted to do the same (i.e. process those images that were panel-corrected by myself in agisoft but then I get a strangely quick processing, with mostly gaps in the orthomosaic).

Did anyone get very good well balanced orthomosaic in Agisoft despite having flown in non-ideal conditions using DJI P4M? If yes, any advice will be appreciated!

Thank you.


############EDIT##############
It is not a matter of panel. With or without panel, the problem is the same.

[bravemaster3]

I just got some better results when using Average blending mode to produce the orthophoto instead of Mosaic mode. Not ideal because you can still see some evidence of not well blended seamlines, but definitely better than the mosaic mode for these datasets.
I am thinking of exporting the orthophotos individually and then I will try mosaicking in QGIS and see if that gives anything better.

I hope this helps anyone who might get into a similar situation.

[bravemaster3]

And here is what I get when I take out orthophotos (radiometrically corrected and orthorectified during the same process in Agisoft... as in the first photo of this post), and mosaic it in Orfeo Toolbox instead.
Only problem, gaps are then not filled with this approach

So, I don't know, but something is off with the blending in Agisoft when illimination conditions are not similar, but as you can see it should not be since the same could be mosaicked well elsewhere.

I hope these help someone that has the same issue.

[nprokofyev]

Hi, how many flights do you have here? Are they processed in single chunk? Please note that every single chunk can only have one set  of calibration parameters (set by reflectance panel). In case you process several flights in one chunk, you will have one properly calibrated flight and malcalibrated others.

[bravemaster3]

Hi, how many flights do you have here? Are they processed in single chunk? Please note that every single chunk can only have one set  of calibration parameters (set by reflectance panel). In case you process several flights in one chunk, you will have one properly calibrated flight and malcalibrated others.

I have several flights, but for reflectance calibration, I calibrate each flight separately. So, I use one or 2 images of a single panel per flight.
Only after the reflectance calibration, I merge the chunks for further processing. I also tried processing chunks separately till the end, but didn't help.

[bravemaster3]

Hi, how many flights do you have here? Are they processed in single chunk? Please note that every single chunk can only have one set  of calibration parameters (set by reflectance panel). In case you process several flights in one chunk, you will have one properly calibrated flight and malcalibrated others.

Now, what intrigues me is that when I use the exported orthophotos to blend them in Orfeo Toolbox, you could see that the mosaic looks good. The mosaic in Agisoft looks what you saw....