Workflow to process the Photography of "PHANTOM 4 RTK"

[Rucolas]

Hello partners:
I am trying to process a set of photographs taken with the Phantom 4 RTK.
The photographs have in their metadata the precisions in XYZ with which they were made.
When I upload the photos in the PhotoScan 1.4.3 does not recognize that data, it gives me the precision that PhotoScan has by default. 10 meters.
I have tried to load the same photos in Pix4D, if you load the precisions of XYZ (attached photograph).
Can someone tell me the workflow to process the photos taken with the Phantom 4 RTK?
Thank you very much for your cooperation and sorry for my English.
Greetings.

[Alex737]

Where can I download photos taken by phantom 4 rtk?

[3DFranz]

I can give it also a try, If you can share the fotos.
Be sure that you have checked (under PhotoScan settings) "load xmp Data..."

[Rucolas]

Where can I download photos taken by phantom 4 rtk?

Attached link to download the file with 18 photographs taken with the Phantom 4 RTK, as well as, all the files generated by the drone.

https://we.tl/t-fQ3g2XPNRG

[Rucolas]

I can give it also a try, If you can share the fotos.
Be sure that you have checked (under PhotoScan settings) "load xmp Data..."

Yes, I have the "load xmp Data ..." option activated
But when I add the photos, the precision it gives me is 10 meters, the standard configuration of Photoscan (attached image)

[wizprod]

Can you post your configured import window?

If I remember correctly, there is also a an option in the Preferences for loading the accuracy from file.

[Rucolas]

Can you post your configured import window?

If I remember correctly, there is also a an option in the Preferences for loading the accuracy from file.

[SAV]

Hi Domingo,

Try to write the EXIF information of your imagery into an external CSV file using EXIFTOOL.
Follow this instruction: https://medium.com/@hunchly/bulk-extracting-exif-metadata-with-hunchly-and-exiftool-164d67c8d7e2

When you have generated the exif.csv file, open PhotoScan, load your images and click on the IMPORT icon in the reference pane (first icon from left). It will allow you to import camera locations (and orientations), not just GCP/marker information. Just make sure that the name in the label column matches the filename of your images.

If you can share your imagery again then I'll test it for you.

All the best.

Regards,
SAV

[Rucolas]

Hi Domingo,

Try to write the EXIF information of your imagery into an external CSV file using EXIFTOOL.
Follow this instruction: https://medium.com/@hunchly/bulk-extracting-exif-metadata-with-hunchly-and-exiftool-164d67c8d7e2

When you have generated the exif.csv file, open PhotoScan, load your images and click on the IMPORT icon in the reference pane (first icon from left). It will allow you to import camera locations (and orientations), not just GCP/marker information. Just make sure that the name in the label column matches the filename of your images.

If you can share your imagery again then I'll test it for you.

All the best.

Regards,
SAV

Thank you very much for your answer SAV

I'm doing the same as you, but from the file with extension * .RMK, generated by the Phantom 4 RTK.

In that file are the EXIF data of all the photographs, I edit it to keep only the data I need and import it following the same method.

My intention is to make PhotoScan automatically read the data of the accuracy of the coordinates as the Pix4d mapper does.
To avoid having to do a previous process before adding the photos in PhotoScan.

I would also like to know if to process some photographs with precision of 1 or 2 cm photocenters, taken with a Phantom 4 RTK, it would be necessary to modify the type parameters that I am using to process the photographs of a Phantom 4 PRO

Receives a warm greeting.

[spiker]

Hello to all! :-)

We have also invested in the DJI P4RTK with hopes of a successfull process with Photoscan but unfortunatly it is a disapointment in regards of camera/lens calibration.
With the DJI P4P we experienced large problems with Photoscan in the optimize process after alignment. To get good surveying data with GCPs all cam stations had to be unchecked and then GPCs clicked, before optimize.
Now we have done several tests with P4RTK and photoscan can NOT process pictures from P4RTK. We have surveyed the same area several times with and without camera distortion, with and without adaptive camera model fitting when aligning.
But compared with earlier DSMs and GCPs we get about 30-50cm off.

Will Photoscan ever fix DJI P4RTK pictures?
Or what have we missed in the process?

Regards

Johan

[Dragline]

I have taken a look at the data and noticed a few other things.

When opening the first image via exif tool, it seems as if the location is listed twice in the metadata. Once in the EXIF section (Red in photo) and once in the XMP section (Blue in photo). Agisoft seems to be loading the EXIF data. This is important as the EXIF data rounds the final 2 digits. This results in ~2cm error. 

Another thing to notice is that accuracy figures (RtkStdLon... etc) are for each axis, whereas Agisoft only has a place for one generalized accuracy value. Perhaps these values need to be combined to one general value for Agisoft to use them?

Location data as visible in Exiftool GUI:

 

Location data as visible in Photoscan:

[Alexey Pasumansky]

Is it possible to send an example of such image from DJI with the accuracy values embed to the XMP meta data? It seems that the namespace where the values are written to is not supported, but it should be possible to implement the support for it quickly.

[Alexey Pasumansky]

We have also invested in the DJI P4RTK with hopes of a successfull process with Photoscan but unfortunatly it is a disapointment in regards of camera/lens calibration.
With the DJI P4P we experienced large problems with Photoscan in the optimize process after alignment. To get good surveying data with GCPs all cam stations had to be unchecked and then GPCs clicked, before optimize.
Now we have done several tests with P4RTK and photoscan can NOT process pictures from P4RTK. We have surveyed the same area several times with and without camera distortion, with and without adaptive camera model fitting when aligning.
But compared with earlier DSMs and GCPs we get about 30-50cm off.

Will Photoscan ever fix DJI P4RTK pictures?
Or what have we missed in the process?

Hello Johan,

If you have some problems with the data processing, maybe you can share the problematic image set with support@agisoft.com, so that we can try to reproduce the issue?

[Rucolas]

Is it possible to send an example of such image from DJI with the accuracy values embed to the XMP meta data? It seems that the namespace where the values are written to is not supported, but it should be possible to implement the support for it quickly.

Attached link to download the file with 18 photographs taken with the Phantom 4 RTK, as well as, all the files generated by the drone.
https://www.dropbox.com/s/b38by1vpj68n0g4/Phantom%204%20RTK.rar?dl=0

[Alexey Pasumansky]

Thank you, Domingo, I have missed the previously shared link.