1
General / RTK Positioning accuracy with Mavic 3E
« on: January 29, 2024, 05:51:24 PM »
Hey everyone,
until a few month ago, i used the Phantom 4 RTK for my aerial surveys.
The system estimated the images location accuracy to be around 1-2cm H / 2-3cm V.
When aligning images, i usually got RMS around 2cm which seemed plausible. The GCPs usually confirmed this.
Now, with the Mavic 3E RTK, i just cannot seem to reach this quality for the RTK data, even if the system itself estimates its own accuracy to be very good, sometimes even sub 1cm (lol, of course DJI)
But when i process the data in metashape, the ugly truth comes out: usually i get around 4-5 cm RMS, with outliers of 20cm errors or more.
Now with the P4, i used an SDK combo with the old GS Pro App, which let the drone stop for taking each image. The Mavic with DJI Pilot 2 just takes images while moving.
I assumed this might be a problem, but when i did some expirments, flying a small area with 4 good quality GCPs, with 3 m/s, 6 m/s, 9 m/s at 40m height above ground, there was no trend of increasing errors in RTK positions. Kinda frustratingly, during these tests, the RMS from alignment was very good, and GCPs also confirmed a pretty solid quality for the RTK positions.
But it was only a relatively small area, around 40-50 images. For larger areas, and especially when using the Mavics oblique capture mode, the RTK positions just seem to be all over the place.
I tried using Metashapes rolling shutter compensation, and it improved the results somewhat, but still nowhere near where they should be.
Does anyone else have any experiences with this? Another possibility would be ppm error from distance to the nearest reference station (I use the public german NTRIP service, Sapos).
But before i go off and doing a thousand tests to get to the bottom of this, i wanted to ask this very knowledgeble community if anyone had any ideas, or similar, or maybe even different experiences?
Please let me know.
Greetings from germany,
Niklas
until a few month ago, i used the Phantom 4 RTK for my aerial surveys.
The system estimated the images location accuracy to be around 1-2cm H / 2-3cm V.
When aligning images, i usually got RMS around 2cm which seemed plausible. The GCPs usually confirmed this.
Now, with the Mavic 3E RTK, i just cannot seem to reach this quality for the RTK data, even if the system itself estimates its own accuracy to be very good, sometimes even sub 1cm (lol, of course DJI)
But when i process the data in metashape, the ugly truth comes out: usually i get around 4-5 cm RMS, with outliers of 20cm errors or more.
Now with the P4, i used an SDK combo with the old GS Pro App, which let the drone stop for taking each image. The Mavic with DJI Pilot 2 just takes images while moving.
I assumed this might be a problem, but when i did some expirments, flying a small area with 4 good quality GCPs, with 3 m/s, 6 m/s, 9 m/s at 40m height above ground, there was no trend of increasing errors in RTK positions. Kinda frustratingly, during these tests, the RMS from alignment was very good, and GCPs also confirmed a pretty solid quality for the RTK positions.
But it was only a relatively small area, around 40-50 images. For larger areas, and especially when using the Mavics oblique capture mode, the RTK positions just seem to be all over the place.
I tried using Metashapes rolling shutter compensation, and it improved the results somewhat, but still nowhere near where they should be.
Does anyone else have any experiences with this? Another possibility would be ppm error from distance to the nearest reference station (I use the public german NTRIP service, Sapos).
But before i go off and doing a thousand tests to get to the bottom of this, i wanted to ask this very knowledgeble community if anyone had any ideas, or similar, or maybe even different experiences?
Please let me know.
Greetings from germany,
Niklas