Matrice 4E RTK height issue

Iain · Tuesday at 1:43 AM

Hi all,
I have recently flown a survey mission with our Matrice 4E. But when I try to add GCP's to the job there is a 47M error between the drone data height and the GCP height I know the GCP's are correct.

The RTK height says it's at 0 this doesn't seem correct to me but I've never noticed it before and it's been fine on previous jobs before this last one which was just after a firmware update.
Any ideas on how to fix this would be appreciated.

Iain · Tuesday at 1:46 AM

Photo of connected RTK for conformation of the 0M height issue

Dave Pitman · 2026-04-15T10:16:35-0400

Looking up the coordinate being reported for the base, it's elevation should be around 58m, not 0.

Iain · 2026-04-15T10:23:51-0400

Dave Pitman said:
Looking up the coordinate being reported for the base, it's elevation should be around 58m, not 0.

Thanks for the reply.

That's my issue I have GPS data for the same area which I know is correct at around the 58m mark but my drone data says it's between 10-11m in height when it's run off in Terra I have never had this issue before and other than a firmware update the day before the flight nothing has changed.

So I'm at a loss as to what's happening

Dave Pitman · 2026-04-15T10:39:22-0400

What is the source of the NTRIP data?

Iain · 2026-04-15T10:57:18-0400

Dave Pitman said:
What is the source of the NTRIP data?

VRS Network subscription the same one used on the GPS data which is fine.

Dave Pitman · 2026-04-15T11:08:40-0400

Do the height values in the image meta data match reality?

Iain · 2026-04-15T11:09:29-0400

Forgive my ignorance but how do you check that

Dave Pitman · 2026-04-15T11:22:42-0400

There are many free apps to view image exif data. But, for simplicity's sake, if you are on windows, you can right click on the image and select PROPERTIES > DETAILS (maybe similar on mac)

Scroll down to the GPS section and it displays the altitude. This should roughly correspond to the ground elevation at your flight location + the altitude of the drone.

Iain · 2026-04-15T12:01:09-0400

Yes that seems to be roughly correct

Dave Pitman · 2026-04-15T12:06:50-0400

If your GCP heights are correct, and the height of your image exif seems correct, then you should be good.

So, what exactly do you mean when you say "there is a 47M error between the drone data height and the GCP height"?

When the drone is at mission height, it should be different than the GCP height which is on the ground?

Iain · 2026-04-15T12:12:08-0400

So when I have ran off the drone data and processed it to OSGB 36 it should be around 57m In level but it's coming out at 10-11m in level hence the 47m average error in real life ground levels and what the drone data is producing.

So it wouldn't find the GCP'S to fit too as they are higher than the drone flight data.

Dave Pitman · 2026-04-15T12:18:29-0400

Sorry, I'm not following your terminology. I suspect you have something set up incorrectly in your processing. If you want to share a partial set of the images and a .csv with your ground control, I'd be happy to take a look at it.

Iain · 2026-04-15T14:37:57-0400

So I loaded the photos into DJI Terra processed them into OSGB 36 output and then added my GCP's which I located via GPS and the GCP heights are 57m ground OS levels and the drone data is coming out at 10m ground in the same coordinate system.

I'll be happy to share a small sample if that's easier

Dave Pitman · 2026-04-15T15:38:17-0400

I don't use Terra, so no help there. But if you just take the fact that your ground shots are good, and the altitude written to the images are good, then, the only place for a great amount of error added has to come from the processing inputs.

You're not mixing up feet and meters on anything, correct?

Have you tried processing and not applying any CS transformation?

What is the CS used by your NTRIP provider?

If you have Dropbox or equiv., load up some images.

Iain · 2026-04-15T15:56:11-0400

Thanks for your assistance so far.
VRS RTK is on OSGB 36 CS

I will upload some photos tomorrow see if that can help

jaja6009 · 2026-04-16T02:12:17-0400

Reading through your post I have to add that I have never used a VRS service so I cannot comment on the 0.00 elevation in the Base/VRS positioning.

In Terra are you using ellipsoid for elevations or are you using orthometric?

Using H= h+N (In the US we use H=h-N since we are under the ellipsoid) where H=Orthometric Elevation, h=Ellipsoid elevation N=Geoid Undulation

We get 56 = 106 + N with N now being -50 meters.

50 meters is around what your discrepancy is correct?
50 meters is the geoid undulation for that location and is used with GNSS coordinates to obtain an orthometric elevation from an ellipsoid elevation.

Usually points are collected in cartesian XYZ that are then reported as lat long with an ellipsoidal elevation.
You then get the lat long and use a geoid tool to get the geoid undulation for that coordinate and then use the formula to solve for H since you will now have both the ellipsoid elevation h and geoid undulation N.

It seems like you collected your control points in orthometric, but your drone images are ellipsoidal.

I am unfamiliar with your datum, but datums are going to work the same way at the end of the day.

In Terra, make sure you choose the correct datums and coordinate systems, both horizontal and vertical for both the images and the control points. I don't really use Terra and I do remember it being a pain in the but to set the images datum and horizontal and vertical systems.

Iain · 2026-04-16T02:22:13-0400

Thank you for your response I think it is something I've done in Terra the more I think about it as never had a problem before with the drone and RTK flights

So I will try messing with settings also what I have tried is moving the GCP'S by the error difference and when I do this they match to within 15mm which is a work around currently as we have scan data to match it too but I want to sort it for the next project as it's knocked my trust in it.

Hope this makes sense

jaja6009 · 2026-04-16T14:30:37-0400

Iain said:
Thank you for your response I think it is something I've done in Terra the more I think about it as never had a problem before with the drone and RTK flights

So I will try messing with settings also what I have tried is moving the GCP'S by the error difference and when I do this they match to within 15mm which is a work around currently as we have scan data to match it too but I want to sort it for the next project as it's knocked my trust in it.

Hope this makes sense

Are your control points check points or ground control points or a combination of both?

When you are taking away the "error" you are basically doing what I mentioned, although not in the correct way. I am unsure about the resolution of the geoid in your area, but in the US, NAVD88, geoid 18 will have a different undulation value for every single image processed so it is most likely that you will be similar.

When drones are receiving corrections they are expecting the elevations to be in an ellipsoid value.
When you collect control points you setup your GNSS rover to collect in the vertical system you choose. You could do orthometric or ellipsoidal.
It looks like you collected your control points in orthometric. I would need to see the control points csv and some images to confirm.

When you are processing in Terra, it looks like you have the control points in orthometric elevation and the images in ellipsoidal, hence the shift.
Look at your control point csv and see if a column has ellipsoidal values and if so reprocess using those values.
Then in Terra output in your desired horizontal projection and choose to output in orthometric elevations using your preferred project vertical datum and geoid.

Agisoft Pro makes this whole process extremely easy. On a single screen you state your project's output datum/coordinate system, image datum/coordinate system and control points datum/coordinate system. And you can easily create combination system's of your own. Pix4Dmapper has its own mess, and Terra from what i remember was something also not easy in my opinion since you entered this on different screens.

All of this is speculation since I don't see the data. Hope this helps.

Dave Pitman · 2026-04-16T15:44:15-0400

jaja6009 said:
It seems like you collected your control points in orthometric, but your drone images are ellipsoidal.

If this is the case it should be noticed in the values just looking at them. The OP stated that the ground values and image values are what they expected. If the ground was ortho and the images ellipsoid, it should be obvious looking at the values as the image values would be too low compared to the ground if the ground was using ortho. But we are kind of shooting in the dark here.