Height problems between coordinates in photos from m300 and Phantom 4 pro v2

Meta4

Moderator
Premium Pilot
Joined
Jan 8, 2014
Messages
15,170
Reaction score
9,105
When I open the exif of the ohantom photo, the absolute height is 218.71 and relative 28.4. how did you find the 31.5
I'm not sure where you are getting those numbers.
i-t9Tcgng.jpg
 
Joined
Apr 23, 2020
Messages
17
Reaction score
1
Age
49
I'm not sure where you are getting those numbers.
i-t9Tcgng.jpg
Meta4 these are from my Phantom 4 Pro V2. I think when he triple posted he might of thought he posted his own Phantom 4 image.

OP if this is the case please repost an image from you Phantom.



I looked at your P1 image. Wow I thought you were going to send an H20 image. I gotta really beg my boss to get us the P1 for our Matrice.


Here is my take and pure opinion.


I got the ellipsoid elevation for the coordinates of the Matrice image. 237.33 meters above the WGS84 ellipsoid.

Now take your relative elevation of 104 meters and add it since you were flying at that altitude.

237.33 + 104 = 341.33 meters above the ellipsoid.

This matches pretty close to 344.5 meters from the Absolute elevation in your metadata.

Were you using RTK on at the time, and if so what datum were you using?

For these images, import into Agisoft as ellipsoid values.

Example:

1661829138546.png



With no Geoid information above this will import as ellipsoid heights.

Now if we can get one of your Phantom images in question please.


From just the information you typed on your Phantom Image of 28.4 Relative and 218.71 Absolute.

I will take the ellipsoid height from the coordinates of the Matrice image since it was taken pretty close if I am understanding what you wrote.

Ellipsoid Height 237.33m and add the relative height of 28.4m we get 265.73m which does not match the Absolute altitude of 218.7m if it was referring to the ellipsoid.
Now lets try to see if this is above sea level. I have the altitude above sea level as 195m. Add your 28.4 meters flight height and we get 223.4 meters which is close to your absolute altitude of 218m

If you are importing both datasets into Agisoft as ellipsoid heights, this is your shift.

You can test this by doing this. Import your Phantom images into Agisoft. Then export the file from the Reference Pane so you can work on them in Excel. Change all of your altitudes to 28.4m + 237.33 = 265.73 and then import back into Agisoft.

See if this gets rid of the shift.
 

Attachments

  • 1661829266710.png
    1661829266710.png
    26 KB · Views: 17
Joined
Apr 23, 2020
Messages
17
Reaction score
1
Age
49
There are two altitude numbers in the metadata for each image file.
The Relative Altitude is the number you see on screen while flying.
Absolute height is the inaccurate approximation of height above sea level.
Agisoft only uses the relative height and completely ignores the absolute height numbers.

The metadata for those two files shows:
P4 pro
Absolute height 31.5
Relative height 56.2

M300
Absolute height 344.5
Relative height 104.2

That's a difference of approx 50 metres (not 120 m).
Were the drones flying at 56 m (P4 pro) and 104 m (M300), like the data indicates?

If you were to fly the drones at the same height, does Agisoft have any problem working with the image files?
When I imported the Matrice image into Agisoft it did not ignore the Absolute elevation.

OP Matrice Image

1661832982750.png



My Phantom 4 Pro V2 image




1661833071720.png


I do not think 344m is referring to Above Sea level even with a generous buffer around the elevation as 195m ASL is the ground in the Matrice image and he was 104m altitude = 299m. This being an ellipsoidal height would be my bet and its very close when you do the math.

As far as my Phantom 4 Pro V2, if I take the 31.53 as an ellipsoidal height it process with a 4 meter inaccuracy in z which is not too bad considering the consumer grade GPS.

Now looking at his image imported it seems like RTK maybe was on and he was in FLOAT. Notice the accuracies. If in FLOAT it would explain the meters differences we are seeing in some of my examples.
 

Meta4

Moderator
Premium Pilot
Joined
Jan 8, 2014
Messages
15,170
Reaction score
9,105
Hello guys, does anyone have an experience with the height difference in coordinates between m300 and Phantom 4 pro? We have automaticly flight mission with UGCS and m300 for orthos and grids. We use the phantom 4 pro v2 for more detailed pictures for 3D reconstructions. But the photos from m300 has always difference in height more than 120 meters against the phantom 4 pro and the result cant link together in agisoft. Because the m300 photos are in correct positions, but the Phantom photos are somewhere under ground. Does anyone have some idea how to fix it? Or what should we change in our workflow? Thx for you reply.
Sorry .. I was confused by two people posting photos and forgot who was the OP.
Here is the relevant metadata from both cameras.
Ignore the upper numbers and just look at the bottom numbers.
i-SRL9XZB-L.jpg


The M300 image shows a height of 104 metres and the P4 pro shows 28 metres.
Looking at the two images, this looks about right.
The M300 was much higher.

The lens used on the M300 was a 35m equivalent of 35mm.
But the Fixed lens of the Phantom has a much wider field of view with a 35mm equivalent focal length of 24 mm.
I'd be surprised if the difference in cameras and lenses isn't the source of your problems.
Agisoft just can't properly stitch images from such different lenses.
Try shooting with just the one drone for both higher and lower images and see if that solves the problem.
 
Joined
Apr 23, 2020
Messages
17
Reaction score
1
Age
49
Last thing I'll add to support my point.

I emailed DJI and here is what they emailed me back.

It went through their normal Support and to their "engineers" and took a week to get.

"Thanks for your patiently waiting.

Here are the answers to the drones you were asking:
Mavic 2 Enterprise Advanced - Ellipsoid
Mavic 2 Enterprise/Zoom - relative altitude
Phantom 4 Pro V2 - relative altitude
Phantom 4 RTK - Ellipsoid
Matrice 300 RTK - Ellipsoid

Aircraft without RTK (including consumer aircraft) only use relative altitude or TOF altitude. When the aircraft actually uses the altitude, it is calculated based on the barometer itself. It is based on the altitude, not ellipsoid height."

I found this to be spot on except for my Phantom 4 Pro V2.

Here is the results of the test I did for a paper I wrote on Public Safety using drones to produce maps without GCPs and/or RTK and then with GCPs.

4 drones were tested with three of them being DJI. A Phantom 4 RTK was also thrown in to show how when you want to map, use the correct tool. The altitudes flown varied due to attempting to get to a desired GSD. (Some GSD's were incorrectly predicted by DJI Pilot.

DroneSet Altitude (ft)Set Altitude (m)Rel AltitudeAbsolute AltitudeGPS Altitude
Mavic 2 Enterprise Advanced17753.9553.8528.1728.1
Mavic 2 Enterprise2106463.967.6767
Phantom 4 Pro V218255.4756.231.5331.5
Ellipsoid height in area = -32
Geoid Height in area= -34
H = h - n
Orthometric Height = Ellipsoid height - Geoid Height
2m average Orthometric height in area


Set altitude is what I commanded DJI Pilot and GS Pro to fly to attempt the desired GSD.

First aircraft Mavic 2 Enterprise Advanced
It flew at 53.95m which matches up real nice with the Rel Altitude of 53.85m.
Absolute Altitude was 28.17m. There is no way possible this is referring to Above Sea Level. ASL would be around 52m with the ground being 2m ASL. Further evidence is that the ellipsoid height in this area is -32m. Add your flight height of 53.85 to this and you get 21.85m which is close enough to 28.17 m to be the inaccuracy of a consumer grade GPS on this drone. It makes much more sense than saying this is ASL. And it also matches what DJI said about reporting as above ellipsoid.

For this to be reported as Above Sea Level it would be off by: around 24m
For this to reported as above ellipsoid it is off by: around 6m



Second aircraft Mavic 2 Enterprise
It flew at 64 meters as set in DJI Pilot which is almost spot on with the Rel Altitude of 63.9m
Absolute altitude is listed as 67.67m. For this to be ASL in this location (2m ASL) it should be around 62m. This is close.
Lets look at if this could be ellipsoid. Add your flight height of 53.85 to -32m (ellipsoid) and you get 31.9 meters. This is again too great of a difference in what it should be. This matches up with what DJI emailed me back again considering the consumer grade GPS on this drone.

For this to be reported as Above Sea Level it would be off by around: 6m
For this to be reported as above the ellipsoid it is off by: around 32m



Last but not least, let's throw in the wild card of this test.
Phantom 4 Pro V2
DJI said that this would not be reported as ellipsoid heights.
It flew at a commanded 55.47m which is pretty close to our 56.2m Rel Altitude.
If this was ASL this should be around 54m for Absolute Altitude. We are not even close with 31.53 being in our metadata.
Okay, lets see if this makes more sense to be reported as ellipsoid. Add our flight height of 56.2m to our ellipsoid value of -32 and we get 24.2.


For this to be reported as Above Sea Level it would be off by around: 22m
For this to be reported as ellipsoid it would be off by around: 7m

Let's add something more concrete.
This Phantom 4 Pro V2 had a PPK kit on it and its dataset was run both with the original non-corrected images and then the PPK tagged images.

The PPK corrected image (Base Station setup on a low Standard Deviation NGS monument that is also a part of hybrid Geoid 18 and used to constrain the new GRAVD gravity model) had a corrected altitude in ellipsoid as 26.09m which matches up decently for our 31.53m when we consider the consumer grade GPS.

Unfortunately the Phantom 4 Pro V2 does not match what I was told by DJI. But it does fit better as an ellipsoid height and when processed in Agisoft with the software expecting the image's altitudes to be above ellipsoid, it processed with a vertical error of 4m.

Summary of RMSE with elevations being reported as follows:
Mavic 2 Enterprise Advanced: Above Ellipsoid
Mavic 2 Enterprise (Zoom): I mass edited the metadata altitude to 32m for all images to reflect the flight height above the -32 ellipsoid, then had Agisoft expect elevations to be referred to ellipsoid
Phantom 4 Pro V2: Above Ellipsoid

Please ignore the other aircraft tested as I copied and pasted from my paper.

sUASXYZTotal
Phantom 4 RTK Double Grid (Oblique 60*)0.77 cm0.50 cm1.71 cm1.94 cm
Phantom 4 RTK Single Grid (1 Diagonal Line Oblique) No GCPs0.66 cm0.50 cm3.08 cm3.19 cm
Phantom 4 Pro V2 PPK2.11 cm1.41 cm10.67 cm10.97 cm
Anafi Thermal0.36 m1.85 m0.34 m1.92 m
Mavic 2 Enterprise (Zoom)1.51 m1.51 m0.74 m2.26 m
Phantom 4 Pro V20.21 m3.23 m4.20 m5.30 m
Mavic 2 Enterprise Adv.4.38 m0.86 m4.52 m6.35 m

Different DJI drones report their elevations differently and it is not always ASL or ellipsoid as reported by both DJI and this one of several tests I did. Pix4D actually caught the Mavic 2 Enterprise Advanced as being above ellipsoid early on when I was suffering from incorrect elevations. I then emailed DJI and go the above response.
 

Meta4

Moderator
Premium Pilot
Joined
Jan 8, 2014
Messages
15,170
Reaction score
9,105
I found this to be spot on except for my Phantom 4 Pro V2.
It will be correct for the Phantom 4 pro V2 as well.
Absolute Altitude was 28.17m. There is no way possible this is referring to Above Sea Level.
You need to forget the Absolute Altitude number shown in metadata.
It's junk and completely useless.

And don't be confused by DJI labelling the data field as GPS altitude - it does not come from GPS.
It's calculated with a very poor algorithm that assumes atmospheric pressure at sea level is a constant (but it's not).
It's commonly out by +/- 200 feet or more.

If you do a little testing you'll find that the same drone will give very different Absolute Altitude numbers flying in the same place on different days.
Don't try to make sense of the number for Absolute Altitude because it's useless junk.
Just ignore it.

Here are a couple of examples from the same drone in the same area on different days:
i-f3jT4GL-M.jpg


i-dJG5SNj-M.jpg





 
  • Like
Reactions: jaja6009
Joined
Apr 23, 2020
Messages
17
Reaction score
1
Age
49
It will be correct for the Phantom 4 pro V2 as well.

You need to forget the Absolute Altitude number shown in metadata.
It's junk and completely useless.

And don't be confused by DJI labelling the data field as GPS altitude - it does not come from GPS.
It's calculated with a very poor algorithm that assumes atmospheric pressure at sea level is a constant (but it's not).
It's commonly out by +/- 200 feet or more.

If you do a little testing you'll find that the same drone will give very different Absolute Altitude numbers flying in the same place on different days.
Don't try to make sense of the number for Absolute Altitude because it's useless junk.
Just ignore it.

Here are a couple of examples from the same drone in the same area on different days:
i-f3jT4GL-M.jpg


i-dJG5SNj-M.jpg

Meta 4, you were correct. Before reading your response I started to think that I had only a small sample size (I do not map without RTK/PPK or GCP unless testing something.)
I increased my sampling size and just as you said, it did not hold up well. And DJI themselves said the Phantom 4 RTK did no report as ellipsoid.


What is your opinion though on DJI saying that for any drone that can have RTK (Phantom 4 RTK, Mavic 2 Enterprise Advanced and Matrice 300 RTK) that the Absolute Altitude for these drones is based on ellipsoid even when not using RTK?
As I look at more data for these drones and my sampling size increases that this is holding up much better.

I will keep on testing this.



OP I would do this.
If the Matrice was using RTK, then you're fine. Although the accuracy assigned to the one image seems like you were in FLOAT.

Next, get the coordinates of the takeoff point and get the ellipsoid height for that spot. Add the height commanded on your mapping software data collector software to the ellipsoid height, then change all of your altitudes in Agisoft to that number. Make sure you import them as ellipsoid (No info in geoid field).
Leave the assumed accuracy as 10m.

This is how I do it from non-RTK and no GCPs and it works well and takes only a couple of extra minutes.
 

Recent Posts

Members online

Forum statistics

Threads
142,310
Messages
1,462,038
Members
104,074
Latest member
Fabitamag