Show Posts

This section allows you to view all posts made by this member. Note that you can only see posts made in areas you currently have access to.

Messages - uarsteam

Pages: [1]
General / Re: RTK + single GCP
« on: February 08, 2022, 09:25:40 PM »

I've recently started working on a project with an RTK setup (DJI Phantom 4 Multispectral + DJI D-RTK 2 base station). My understanding is that this sort of RTK positioning will be centimetre-precise and internally consistent, but may be offset in real world coordinates by some decimetres.

I have read that this uniform offset can be corrected using a single GCP (e.g. This would be great because we are operating in an environment where, due to time and access limitations, we can only place a single GCP.

My plan was to use the RTK camera positions during image alignment to obtain a precise geometry, and then introduce the single GCP during the camera optimisation process. Effectively, I want perform a translation on my dataset to align it with the single GCP.

However, after going through the full Metashape processing chain it appears that the GCP marker still reports nearly a metre of error, which I can see when I export the orthophoto and view it in GIS software. After searching this forum, I'm thinking my problem may be that I do not uncheck the photos in the Reference window before performing an optimisation, leaving Agisoft to try and model both the camera positions and GCP position together. Is this the case? If so, my new processing chain would look as follows:

  • Import photos (and RTK camera positions)
  • Align photos
  • Import GCP and mark in photos.
  • UNCHECK all photo locations. < I currently do not this
  • Optimise cameras
  • Dense cloud, DEM, ortho, etc.

Is this the case? Or would unchecking photo locations before performing an optimisation mean that I 'lose' the benefits of drone RTK positioning?

Thanks in advance!

Hey tom_rc33,

I've meddled with using or dropping image geotags in addition to incorporating GCP markers for additional spatial refinement. In my experience, there are differences, but they can be pretty modest. Your post got me wondering: how did you survey the location of your GCP, and did you survey the location of the base station? As mentioned, DJI's base station will have a global accuracy of about 1-2 m, but RTK should still offer a local accuracy of approx. 1-2 cm.

General / Re: Radiometric Calibration in version 1.4.4 build 6848
« on: February 08, 2022, 08:30:42 PM »
Was a solution ever found to this question? I have experienced similar issues during the calibration process. Thanks!


Is anyone else having trouble understanding the reflectance observed for NIR band images. There appears to be a significant brightness artifact located somewhat centrally to the image with a considerable drop-off as you approach the periphery of the image. It sort of follows the shape of a vignette process, however the brightness doesn't appear to be systematic from image to image. I've received no assistance from the manufacturer on the topic. Any thoughts whatsoever are greatly appreciated. Compressed JPG (not orig TIF) attached for your convenience.


General / Re: Phantom 4 Multispectral
« on: December 20, 2021, 08:56:23 PM »
Sometimes the calibration panel works well but other times not so much. For me, it mostly impacts some blue band captures.

One big issue is that the P4M has a different definition of vignette correction in comparison to Micasense so DJI is discussing this and trying to find the best way to solve the issue.

Given the vignette issue, I suspect that some other metadata values may not have a one-to-one correspondence with MicaSense metadata. Also, given that vignette correction is not successful, and it is part of the calibration panel routine, the calibration panel would need to be in the center of the frame so it isn't in a dark spot in the image. I'm not sure if anything else in the calibration routine "breaks" due to the vignette issue. The vignette can be solved with "calibrate colors" later on in processing, but the DJI metadata parameters might need to be translated to MicaSense equivalent, or the Agisoft code might need to be modified for the P4M.

Also, for troubleshooting this, I recommend against trying to use the sun sensor and just stick to panel calibration.


Any update on DJI's deliberation over the vignette definition issue? I recently started working with the P4M data in a big way and noticed an awful "brightness" or "hotspot" effect in roughly the center of low-altitude images. It's especially noticeable in the NIR band images, resulting in images where vegetation reflectance appears completely saturated in the center of the image but fades as you move away from the center. Is this to what you are referring about DJI's "discussions", or is it some other phenomenon?



Pages: [1]