Hi all

  

We have a problem with our Zenmuse L1.

  

We realized that if we don't use GCPs, the reconstruction with DJI Terra Pro has a vertical error of over 40 meters.

  

Does anyone have experience with a similar problem?

Is it possible to do something without sending the Zenmuse for assistance?

There is a very good chance that the problem comes from the things between the chair and the keyboard.  
Do you have an idea of the vertical datum you use?  Aren't you trying to fit a WGS84 into an ITALGEO95?

LV_Forestry Posted at 1-20 08:20
There is a very good chance that the problem comes from the things between the chair and the keyboard.  
Do you have an idea of the vertical datum you use?  Aren't you trying to fit a WGS84 into an ITALGEO95?

Thank you for your trust... and for the insult...
I thought I could ask on a technical forum and not find myself the classic social media expert...
No, I'm not trying to fit into an ITALGEO95
I have been using the Zenmuse L1 since November 2021, I would not have opened a post asking for help without having verified the data acquired in the past and that I had not made a mistake in the procedure.
Thanks anyway

saimon66 Posted at 1-20 10:12
Thank you for your trust... and for the insult...
I thought I could ask on a technical forum and not find myself the classic social media expert...
No, I'm not trying to fit into an ITALGEO95

There is no insult, it is a common mistake. The user thinks to choose the right vertical datum in Terra because they are selectable, unfortunately they are not implemented.

40m... it's still weird because it's the average difference between WGS84 and ITALGEO95.

Make sure you have captured your GCPs in WGS84 to match what comes out of TERRA.


Feel free to send a dataset for further check

Just like that, no offense, the value of ITAGEO90 at your address is 39.9m...

LV_Forestry Posted at 1-20 11:36
Just like that, no offense, the value of ITAGEO90 at your address is 39.9m...

[view_image]

You are insisting on a coordinate error.

I make a recap of the latest tests performed with more details:
- RTK flight with NTRIP data acquisition
- With the same NTRIP connection I collect data with the rover before the flight
- Both flight data (verified with Airdata) and GCP data are perfectly aligned: WGS84 UTM32N
So, for me, IMU Matrice 300 and rover return correct and consistent data.
- The reconstruction in DJI Terra is imposed with WGS84 UTM32N: the heights of the take-off points verified in the LAS are different from the drone data with values ranging from 32 meters to 54 meters
This doesn't seem normal to me.
I have never changed anything in my procedures: any changes to the coordinate system I perform from programs external to DJI Terra.
I would be happy if the error was mine and you were right that the difference in height is attributable to the use of ITALGEO, but it is inexplicable to me to have different height differences on different processes.
I have already opened a ticket in DJI: it could be a problem with the Zenmuse IMU...

saimon66 Posted at 1-21 02:12
You are insisting on a coordinate error.

I make a recap of the latest tests performed with more details:

Please, in Terra choose Geoid EGM96 then perform reconstruction and see if the error persist.

LV_Forestry Posted at 1-21 02:32
Please, in Terra choose Geoid EGM96 then perform reconstruction and see if the error persist.
[view_image]

I have already done this test:
- Takeoff altitude: 77 meters (from Airdata for confirmation)
- Height from DJI Terra: 121 meters
- Height by entering EGM96: 108 meters
Unfortunately it has no logic.
I was chatting a little while ago with DJI and they asked me for more data to verify the problem: they will give me their answer in a few days.
There is a possibility that the L1 is faulty...
Also because it has been used many times in the past.
We have the DJI Care Refresh…
The problem as usual is shipping and waiting to get it back...
I'll update the post in a few days as soon as I have answers from DJI.

saimon66 Posted at 1-21 03:00
I have already done this test:
- Takeoff altitude: 77 meters (from Airdata for confirmation)
- Height from DJI Terra: 121 meters

Could you please provide take off point coordinates ?
Air data is not a GIS approved source

saimon66 Posted at 1-21 02:12
You are insisting on a coordinate error.

I make a recap of the latest tests performed with more details:

"RTK flight with NTRIP data acquisition / With the same NTRIP connection I collect data with the rover before the flight"


I believe that rover part you mean GNSS RS2 antenna. It is not because you are on the same NTRIP network that the vertical datum selected is the same.


"Both flight data (verified with Airdata) and GCP data are perfectly aligned: WGS84 UTM32N"


UTM32N is not a vertical datum !
I just took a look at AirData, the altitudes come from a generic global model, it is in no way to be taken as something reliable.
If you want valid models, I invite you to go here: RNDT - Home (geodati.gov.it)

I have already sent the data to DJI and they will help me.
Sorry, but it's impossible for me to continue following you: I would have to share all the data and I can't do so for privacy reasons.
And I should explain better, I know.
For example, when I talk about verification with Airdata I mean that the data logs are uploaded to Airdata and I am actually reading the DJI data of the flight: I expect it to be accurate.
The concept is: why did I use the same tools and do the same actions before I had correct data and now I didn't?
Have you experienced a similar error?
I would say no because you insist on possible operational errors.
So, excuse me, but with all due respect to you I await a response from DJI regarding the Zenmuse L1.
Have a nice day.

saimon66 Posted at 1-21 03:58
I have already sent the data to DJI and they will help me.
Sorry, but it's impossible for me to continue following you: I would have to share all the data and I can't do so for privacy reasons.
And I should explain better, I know.

I have had this problem multiple times when I change the vertical datum of the GNSS receiver to mark GCPs from one project to another, and I forget to change it again.  

You sent your dataset to DJI and you don't want to share it with me for privacy reasons... well,...

Check this before sending your L1 :


And dont be fooled by DJI, if its a warranty case, do not ask a care refresh!

Well Saimon, i'm sorry but LV_forestry wasn't insulting you, a wrong vertical datum is the first think i've thought me too, and the error that is exactly the difference between ellipsoid and geoid model looks pretty weird.
Do you use the new DJI Terra 4.0 to process L1 data?

Panph Posted at 1-22 01:47
Well Saimon, i'm sorry but LV_forestry wasn't insulting you, a wrong vertical datum is the first think i've thought me too, and the error that is exactly the difference between ellipsoid and geoid model looks pretty weird.
Do you use the new DJI Terra 4.0 to process L1 data?

Yes
4.01 now

saimon66 Posted at 1-22 05:02
Yes
4.01 now

I've used too the 4.0 version  few days ago with L1 and everything was fine, i use to process with ellipsoidal height with vertical offset.
If i were you, i'd try to process with no geoid or vertical offset, then trying to revert to 3.9.4 and see if there's any differences. Maybe a bug in new version, if the only change in your usual workflow is the new dji Terra.