Hi,

I have been using the DJI Pilot "Mission Flight" feature for mapping. I've noticed when flying the drone never flies at the set altitude, however, when I go to process the data after the flight xref data does not appear to reflect the altitude that the flight was actually flown at. As a result the resulting surfaces appear to be coming in high (If I set the drone to fly at 120m and it flies at 116m the flight seems to come in 4 meters high). Virtually all of the areas I fly require using multiple batteries and each time I take off the drone will fly at a different incorrect elevation. I don't think that processing the flights using only control points and not RTK data or processing each portion of the flight seperately to apply a different correction are adequate solutions for a drone this expensive. Is there a way to either force the drone to fly at the correct elevation or to retreive the actual flight elevation?

Any help would be appreciated

Hi,

so you enter 120m int the mission mode, the 116m are refered to the on screen displayed height or is that something from the excif/xmp data? I have read the dji computes the AGL height as a combination from GPS and Barometer readings. I mean it could be that you have some kind of systematical error in your barometer readings, but Iam just guessing here.... Is the difference always about 4 meters?

" As a result the resulting surfaces appear to be coming in high...." Compared to what? Are you comparing the data with something which is properly georeferenced? In that case keep in mind that the absolut accuracy of you RTK solution (camera postions) is within meters (using no GCPs).

However, I would not recommend flying without any GCPs at all. We usually fly with the RTK solution wich gives a tremendous advantage regarding relative accuracy and also use a couple of ground controlpoints (<10)  for absolute accuracy.

Best regards...
Felix

Felix,

The difference is inconsistent, it changes each time I replace a battery and continue the flight. 116m was the alititude reported for one portion of the flight, I believe the other portions we're flown at 112m and 114m. When I look at the altitude stored in the picture data all but a few (<5) of the over 700 photos claimed to be within 0.5m of the programmed altitude.  The difference in what we set the drone to fly at and what the screen display shows during flight seems to be the same as the error we are having in the final surface.

I work in a mine and each month we have a contractor come and provide a drone survey so I have these as a reference (checking areas that haven't changed) . We were having an issue with our surface warping badly around the edges of the flight because the RTK elevation is significantly different than the control points. We utilize the same Geoid model for survey as we do for processing our flights (we use agisoft). We can obtain good results using GCPs and not utilizing the RTK of our drone, but if that's the best we can do why pay for RTK?  

Are you using the mobile RTK Base station with your drone or do you use the network GPS option of your drone (by that I mean some kind of service provider for DGPS)?

Here are some scenarios:
If you use the mobile RTK base station and you put that somewhere on the field then the resulting absolut accuracy will  be within meters  (and the height will be even worse by factor 2-3 in comparison to the horizontal position components). All your camera positions will relatively still have a high accuracy. So the the camera compound consisting of lets say 700 camera positions, this whole block in itself, will be very accurate. But this whole block (and the resulting surface derived from agisoft) will be shifted in x-y-z direction (because of the lower absolut accuracy). The reason behind this is because the mobile base station receives a GPS position as a single receiver and your RTK solution (camera position) is a difference to that inaccurate base station position.
Example: Basestation coordinates B,L,200m (GPS height above reference Ellipsoid). But those coordinates are inaccurate (single GPS Receiver solution). Now the drone is linked to that mobile base station (RTK). Then a camera position will be very accurate (cm-level) with respect to that base station, lets say camera position  is Bc,Lc,280m (above reference ellipsoid). This 280m height from the camera has the same low absolute accuracy. Now in Agisoft you transform this camera position into your coordinate system with your geoid. So your geoid grid file (in Agisoft) uses Bc and Lc to interpolate the geoid undulation and you end up with h (above geoid) = Hc (280) - geoid undulation, but if Hc has an absolut accuracy within meters then your final height will also be off by a couple of meters and the same applies to your derived surface.
Edit: So if you have a modelled surface that way and you have some kind of another reference surface, which is the result of other survey work and you use the same geoid model, then it is still ossible end up with a difference in meter level.

Edit: Actually you also have the option to set up the mobile base station over a known point. I have read that this still causes problems sometimes. I have not tested this yet.

Network RTK solution:
Things look different when you use a network service for your drone as an RTK solution (so no mobile RTK base station). Here in Germany we have a service called SAPOS which consists of fixed base stations evenly distributed through the country. As a user you can connect your drone to that service and receive a very high relative and absolute accuracy, because those fixed base stations are very accurate. So with that solution you should be "safe".

You mentioned "normal survey work with no drones", so I assume you also work with GPS receivers connected to some sort of network DGPS correction service. Take that receiver, survey a point, note the corrected position and compare that to the coordinates of the mobile RTK base station (DJI) set up over the same point. If you have problems to read out the position of the mobile RTK base station (DJI) then you could also position the mobile RTK base station (DJI) somewhere else and put the linked drone over that point, take a picture and compare that GPS position. There will be a difference within meters.

Just a comment to Agisoft (also using it with the Matrice 210RTK): The introduction of GCPs does not mean that Agisoft will not use the RTK data. This will all go together into the bundle adjustment. So RTK is not a waste at all.

The major problem with DJI is that it is all black box and most of their systems are so badly documented.

By the way, I dont care at all of the height above ground based on the barometer (the one displayed on screen), because I dont use that height for calculations. This height is very inaccurate as it is very dependend on temperature wind etc. The only use it has for me is to calculate the expecting GSD (Ground Sampling Distance).

Best regards,
Felix