Hello,
I am trying to do a simple orthophoto just to test how is the drone handling it. I'm facing a strange issue.
I'm at an elevation of 5m above sea level. I'm starting a flight plan mission that is at an altitude 100m. So photos should with elevation tag of 105m.
Instead of this, I'm getting photos with an elevation tag 35m. And this is happening almost every time.
The difference between actual elevation and captured elevation is around 50-60-70m. I'm checking this info in the picture properties in windows.
Do you have any idea how to fix it?
Thanks!

I am trying to do a simple orthophoto just to test how is the drone handling it. I'm facing a strange issue.
I'm at an elevation of 5m above sea level. I'm starting a flight plan mission that is at an altitude 100m. So photos should with elevation tag of 105m.
Your drone doesn't have any way to tell how high the drone is above sea level.
In your example the height in your image metadata would be 100 metres.

Instead of this, I'm getting photos with an elevation tag 35m. And this is happening almost every time.
The difference between actual elevation and captured elevation is around 50-60-70m. I'm checking this info in the picture properties in windows.
Your photos have two different altitude numbers in their metadata.
One is Relative Altitude - height above launch point as seen on your screen.
The other is Absolute Altitude and it's incorrect and can't be used for anything.
WIndows properties is just showing one of those and it sounds like it's showing the Absolute Altitude number.

Do you have any idea how to fix it?
For photogrammetry you shouldn't need to fix anything.
Photogrammetry software will automatically use the Relative Altitude data.
If you upload a full-sized original jpg to Dropbox or similar and post a link, I can show you the two different height numbers.

Hi there. Thank you for reaching out. The photo altitude information is obtained after combining the barometer, GPS (with a strong GPS signal), and sensor data, which is affected by temperature and air pressure. DJI cannot ensure that it is the same as the actual GPS height above sea level. If you need a detailed altitude, check the height above sea level based on the GPS longitude and latitude. Use the KUSO Exif Viewer to check the relative altitude value. The sum of these two values is the actual altitude of the aircraft or you may purchase an aircraft with an RTK. We appreciate your understanding and support.

https://we.tl/t-XdM45YeBLg

In this image elevation should be 245.2m
Captured elevation above sea hight level is 182.1m
I'm not in flight, so relative altitude is 0.
Again difference is more than 60m.

djiuser_74NfVsxgonye Posted at 5-16 23:35
https://we.tl/t-XdM45YeBLg

In this image elevation should be 245.2m

Forget DJI's Absolute Altitude because it's totally useless.
It's usually wrong and the error can be +/- 200 metres or more.
The error will vary with the air pressure on the day you took the photo.
The photogrammetry software that I've used ignores it and uses the relative altitude.

If you do test shots from the same place in different weather, you can see how it gives very different Absolute Altitude numbers.


Here's the relevant section of the image metadata:

djiuser_74NfVsxgonye Posted at 5-16 23:35
https://we.tl/t-XdM45YeBLg

In this image elevation should be 245.2m

There is a gap of 100m +/-, which is not so bad
In principle if you put the drone in the air (above buildings) you should gain in precision.

LV_Forestry Posted at 5-17 07:42
There is a gap of 100m +/-, which is not so bad
In principle if you put the drone in the air (above buildings) you should gain in precision.

That makes no difference.
If that was true, I wouldn't be seeing fluctuations in Absolute Altitude of more than 100 metres when flying only 10 metres above the sea, miles from land.
This week I shot one photo with an absolute altitude of -147 metres, but the drone was 10 metres above the sea.

DJI's absolute altitude is junk.
Fortunately photogrammetry software doesn't use it.

Labroides Posted at 5-17 14:38
That makes no difference.
If that was true, I wouldn't be seeing fluctuations in Absolute Altitude of more than 100 metres when flying only 10 metres above the sea, miles from land.
This week I shot one photo with an absolute altitude of -147 metres, but the drone was 10 metres above the sea.

I strongly suspect that the "gpsfusionalt" feature was introduced deliberately to handicap the possibility of doing 3D photogrammetry.  

With the Mavic 2 generation there is no this crap, and the altitude is rather accurate.

Labroides Posted at 5-17 14:38
That makes no difference.
If that was true, I wouldn't be seeing fluctuations in Absolute Altitude of more than 100 metres when flying only 10 metres above the sea, miles from land.
This week I shot one photo with an absolute altitude of -147 metres, but the drone was 10 metres above the sea.

What is the absolute altitude

USER001 Posted at 5-18 03:32
What is the absolute altitude

It's nothing ... it's useless junk.

LV_Forestry Posted at 5-18 00:26
I strongly suspect that the "gpsfusionalt" feature was introduced deliberately to handicap the possibility of doing 3D photogrammetry.  

With the Mavic 2 generation there is no this crap, and the altitude is rather accurate.

I strongly suspect that the "gpsfusionalt" feature was introduced deliberately to handicap the possibility of doing 3D photogrammetry.  
Since it makes no difference to photogrammetry, I strongly suspect you are incorrect.
Although the parameter label IS GPSfusion, I don't believe it uses GPS at all.

With the Mavic 2 generation there is no this crap, and the altitude is rather accurate.
I don't have a Mavic 2 to check, but the Phantom 4 pro which came out before the Mavic 2 had the same Absolute Altitude issue as the current models and is an excellent mapping drone, so I'm not convinced.

USER001 Posted at 5-18 03:32
What is the absolute altitude

The absolute altitude in this case is the reference to the WGS84 ellipsoid. Which is the "zero level" of the reference system used by what is commonly called GPS.

Take a look at this if you want to know more:
https://forum.dji.com/forum.php?mod=viewthread&tid=305829
This makes perfect sense because to do 2D photogrammetry it is actually not a problem if the error is several hundred meters, provided that it is constant. On the other hand, for 3D it is more problematic if the need is to georeference the scene with precision.

Labroides Posted at 5-18 04:52
I strongly suspect that the "gpsfusionalt" feature was introduced deliberately to handicap the possibility of doing 3D photogrammetry.  
Since it makes no difference to photogrammetry, I strongly suspect you are incorrect.
Although the parameter label IS GPSfusion, I don't believe it uses GPS at all.

I can confirm that the photogrammetry software Agisoft Metashape and PIX4D at least, use the absolute altitude contained in the EXIF.
See the example below with the photo shared by our friend the OP:


This is a very important parameter because photogrammetry is not just 2D.


And I also confirm that the model still used by our friend's drone is "GpsFusionAlt"


For P4P I confirm that it is the same misery as the others. Example below, the absolute altitude is -31m, it should be 80m. Most of the time it is correct.

LV_Forestry Posted at 5-18 06:18
The absolute altitude in this case is the reference to the WGS84 ellipsoid. Which is the "zero level" of the reference system used by what is commonly called GPS.

Take a look at this if you want to know more:

Interesting. Thanks for the information.

Labroides Posted at 5-17 14:38
That makes no difference.
If that was true, I wouldn't be seeing fluctuations in Absolute Altitude of more than 100 metres when flying only 10 metres above the sea, miles from land.
This week I shot one photo with an absolute altitude of -147 metres, but the drone was 10 metres above the sea.

Can you show us the photo?

Labroides Posted at 5-18 04:46
It's nothing ... it's useless junk.

Not a very helpful answer. Don't worry, I found out now.

LV_Forestry Posted at 5-18 06:18
The absolute altitude in this case is the reference to the WGS84 ellipsoid. Which is the "zero level" of the reference system used by what is commonly called GPS.

Take a look at this if you want to know more:

Now all you have to do is explain how it varies so much for the same point on different days.
I've shown you the evidence, but am still waiting for you to explain that away.

USER001 Posted at 5-18 07:02
Not a very helpful answer. Don't worry, I found out now.

Correction ... it's an accurate and very useful answer.

Labroides Posted at 5-18 07:23
Now all you have to do is explain how it varies so much for the same point on different days.
I've shown you the evidence, but am still waiting for you to explain that away.

But what is the connection? You're just trying to generate yet another drama.

The point is very clear, there are Mavic2 drones and the enterprise range which works like a "Garmin" GPS and which indicates an acurate pure GNSS altitude (+/- 20m without RTK).
And other drones that use this “Fusion” mode described as calculating altitude based on GNSS and barometer which gives totaly useless results. It's DJI who describes it like this, not me.

With P4P I see exactly the same thingas you. There are days when things go well, others not at all. For what ? This question must be answered by someone at DJI. So call "the relevant team".

Labroides Posted at 5-18 07:24
Correction ... it's an accurate and very useful answer.

It's nothing...
This is a not a good way to start an answer by any means.
it's useless junk.
You are not even telling me what it is, so I don't even know if it is useless junk.
Why don't you be a bit more helpful around the forum?
Correction ... it's an accurate and very useful answer.
Oh, so now you're a liar?
You are certainly a very good man.

LV_Forestry Posted at 5-18 07:33
But what is the connection? You're just trying to generate yet another drama.

The point is very clear, there are Mavic2 drones and the enterprise range which works like a "Garmin" GPS and which indicates an acurate pure GNSS altitude (+/- 20m without RTK).

You're just trying to generate yet another drama.
It's not another "drama".
It's the same puzzling thing that you've made several unsuccessful attempts to explain.

The point is very clear, there are Mavic2 drones and the enterprise range which works like a "Garmin" GPS and which indicates an acurate pure GNSS altitude (+/- 20m without RTK).
I don't have a Mavic 2 to check, but can't imagine why or how DJI got it right with one model that was released six years ago, but every other consumer DJI drone before and since has a completely useless Absolute Altitude.
My initial reaction to your sugestion is to disbelieve you and ask for some Mavic 2 photos to prove what you say.

And other drones that use this “Fusion” mode described as calculating altitude based on GNSS and barometer which gives totaly useless results. It's DJI who describes it like this, not me.
Was DJi accurate to suggest that the absolute altitude was derived from GPS Fusion?
Would GPS altitude vary by >100 metres?
Relative altitude doesn't ... so where does teh variable error come from?

This question must be answered by someone at DJI. So call "the relevant team".
DJI rarely answer anything and when they do, it's not always accurate.
Mostly it's the users who have to work things out for themselves.

Labroides Posted at 5-18 14:26
You're just trying to generate yet another drama.
It's not another "drama".
It's the same puzzling thing that you've made several unsuccessful attempts to explain.

I made a small selection of photos taken with an M2, in different places in the country.
I use the following formula:
(Absolute altitude) - (realtive Altitude) = "theoretical ground elevation"

I take the raster value from the national geographic agency. I do (raster value) - (theoretical ground elevation)= "altitude difference"
At worst there is a maximal error of approx 25m including the geoid.

I'll let you enjoy the result :





If you want to do the test yourself to check, the raster is here:
Digitālais augstuma modelis | Latvijas Ģeotelpiskās informācijas aģentūra (lgia.gov.lv)

And don't hesitate to ask for photos, I have a 7TB hard drive full of them.

Then, why would DJI have introduced this feature?
I know this won't make you happy at all because I know your point of view on this, we've already talked about it...
This is for anyone who wants to test it, calibrate the "compass" before the flight, and see if that changes anything.

Labroides, I know, the compass has nothing to do with GPS...
Aim ! since our last conversation I had the opportunity to integrate a payload on a DJI drone, and before putting it in the air I thought of you, and I said to myself, is that I won't leave the compass with its factory setting to see what it does? Since I have access to all the RAW values. I mean I can see what goes into the Kalman filter, and what comes out.
Well without calibrating the compass here is what it gives as output:




Here is the "real" trajectory from the drone:


After compass calibration:

"The colors are only there to indicate whether the payload records data or not. Its a different field than the previous example"

So there you go, please, don't get involved in drama as usual. At worst don't answer, at best do the test, calibrate the compass, do the flight and share the result with us to see if it has any effect on what comes out of the drone's Kalman filter (data in EXIF). Phantom 4 and other drones causing problems.

There's nothing puzzling about what I bring. Everything is based on facts. I am willing to provide you with all the RAW files you need to help you understand the above. Just ask!

LV_Forestry Posted at 5-19 00:11
I made a small selection of photos taken with an M2, in different places in the country.
I use the following formula:
(Absolute altitude) - (realtive Altitude) = "theoretical ground elevation"

Sorry ... I can't understand what you are saying in the second half of that post.

Labroides Posted at 5-19 02:24
Sorry ... I can't understand what you are saying in the second half of that post.

I suggest that drone owners presenting aberrations in the absolute altitude value recorded in the EXIF, to calibrate the compass before the flight.  
Then to see if this provides the beginning of a solution or not.

LV_Forestry Posted at 5-19 07:23
I suggest that drone owners presenting aberrations in the absolute altitude value recorded in the EXIF, to calibrate the compass before the flight.  
Then to see if this provides the beginning of a solution or not.

I managed to catch that much (but can't see how it would make any difference).
I couldn't make out what you were showing in the illustrations.

Labroides Posted at 5-19 07:36
I managed to catch that much (but can't see how it would make any difference).
I couldn't make out what you were showing in the illustrations.

It is the output of a Kalman filter, GNSS+Magnetometer (compass) which is totally wrong.  
Calibrating the compass completely resolved the problem.

The admins here often claim that the compass data is merged with that of the GNSS, this is certainly what they call "fusion".  

If this is the case then a compass calibration should be able to solve the OP's original problem.

This is just a suggestion, nothing is certain.  This is why I invite all users who have this problem of false altitude to try calibrating the compass.

LV_Forestry Posted at 5-19 08:25
It is the output of a Kalman filter, GNSS+Magnetometer (compass) which is totally wrong.  
Calibrating the compass completely resolved the problem.

Given your record on understanding compass function, I have difficulty accepting your suggestion that the compass has any input to height computation.
That a moderator here might have also suggested something along those lines doesn't give me any confidence either.
They have no technical knowledge at all and can't be relied on for accurate information.

Labroides Posted at 5-19 15:23
Given your record on understanding compass function, I have difficulty accepting your suggestion that the compass has any input to height computation.
That a moderator here might have also suggested something along those lines doesn't give me any confidence either.
They have no technical knowledge at all and can't be relied on for accurate information.

Here we are, I didn't expect more from you.  

If someone else less narrow-minded wants to try it, that would at least make progress on the issue.

LV_Forestry Posted at 5-19 20:32
Here we are, I didn't expect more from you.  

If someone else less narrow-minded wants to try it, that would at least make progress on the issue.

someone else less narrow-minded ??
Says the guy who thinks that compass calibration is related to geographic location, and won't believe the facts about the matter.

Labroides Posted at 5-19 20:59
someone else less narrow-minded ??
Says the guy who thinks that compass calibration is related to geographic location, and won't believe the facts about the matter.

If you have nothing else serious to bring to the conversation, please do not question my words and my ability to understand.

By something serious, I mean things like the following:

Attitude & Heading Reference System (AHRS) (sbg-systems.com)


Learn how an Attitude and Heading Reference System (AHRS) works · VectorNav


Sensors | Free Full-Text | Accuracy Improvement of Attitude Determination Systems Using EKF-Based Error Prediction Filter and PI Controller (mdpi.com)


And finally I refer you to the experience made in post #23, once again, if you want the RAW files to check for yourself, no problem, I can share them.
You see, concrete information, quality sources... it's something other than forum blah blah. isnt it ? Please do not respond until you have tried to see if compass calibration improves altitude definition.

Pix4D will work no matter if Absolute or Relative....
I work with it.
Also for measurements you can use GCP's, one on ground and one at base and 2m height of building...will correct info and give you a very good measurement, like a few cm error.
For an Air3 that is not made for this, is OK.
If you need photogrametry, take an Autel Evo2 Pro V1...is making 2cm accuracy 3D models without GCP's, without extra base.
Cheers.

Geo_Drone Posted at 5-19 22:59
Pix4D will work no matter if Absolute or Relative....
I work with it.
Also for measurements you can use GCP's, one on ground and one at base and 2m height of building...will correct info and give you a very good measurement, like a few cm error.

It will work, we all agree on that.  
What will happen to the accuracy of geo referencing?  

Autel Evo, 2cm... you are confusing it with the commercial specifications of its vision position system.  Extremely misleading information.

LV_Forestry Posted at 5-19 22:04
If you have nothing else serious to bring to the conversation, please do not question my words and my ability to understand.

By something serious, I mean things like the following:

You see, concrete information, quality sources...
Yes .. I see what you posted above, but nothing remotely related to what's being discussed.
I also see nothing at all in it to suggest any connection between the compass and altitude measurement.
Nothing at all.
It leaves me wondering if you misread attitude to say altitude?
The word attitude appears seven times in your post, but altitude isn't mentioned at all.

That's as close as I can find to anything to do with altitude in what you posted.

Labroides Posted at 5-19 23:32
You see, concrete information, quality sources...
Yes .. I see what you posted above, but nothing remotely related to what's being discussed.
I also see nothing at all in it to suggest any connection between the compass and altitude measurement.

No there is no error.  the aTtitude is what is defined by the sensors.  This stabilizes the drone and makes it respond correctly to operator input.  

This aTtitude is defined by numerical values.  It stabilizes the reading on the three axes.  
The hypothesis is that if the starting conditions are bad when initializing the drone, then these digital data will be erroneous, but will remain constant, on the three axes including Z (aLtitude).  

I'm not trying to say that I'm absolutely right.  I currently don't have a consumer drone on hand to test.

LV_Forestry Posted at 5-20 00:07
No there is no error.  the aTtitude is what is defined by the sensors.  This stabilizes the drone and makes it respond correctly to operator input.  

This aTtitude is defined by numerical values.  It stabilizes the reading on the three axes.  

I'll see if I can get around to a little testing in the near future.
But until then, I'm pretty sure that you have no idea what you are talking about.
For one, you don't understand what compass calibration actually does or that it doesn't do anything to the compass or that it doesn't change.
For two, there is no link between the compass sensor and height .. none at all.

Failing to understand anything about that is the only way you might imagine that recalibrating the compass will have any effect on DJI's flawed Absolute Altitude.

Labroides Posted at 5-20 01:06
I'll see if I can get around to a little testing in the near future.
But until then, I'm pretty sure that you have no idea what you are talking about.
For one, you don't understand what compass calibration actually does or that it doesn't do anything to the compass or that it doesn't change.

I understand your doubts. On the other hand I am currently in nature in the middle of nowhere 170km from my last survey, and the camera I use gives totally wrong results. It's not photogrammetry, it's a pushbroom camera. The slightest error in geographic positioning results in nothing more than works of art graphics.

I calibrate the compass, and it's fantastic, everything comes back to normal.

I have access to the RAW data, I'll let you appreciate the deviation before and after calibration of the compass (magnetometer):


And to prove to you that it's real, that it's not something I saw on the internet:


And once again, I'm not saying that this is what will 100% solve the problem expressed by the OP. I say that in my case, for this payload in particular, the equipment manufacturer clearly indicates that the calibration must be done every 50km, otherwise you have to expect crappy datasets.
If you could give a try it would be really cool. Especially since you often fly close to sea level which makes the data even easier to verify.

LV_Forestry Posted at 5-19 23:18
It will work, we all agree on that.  
What will happen to the accuracy of geo referencing?  

Geo ref. will suck on this type of drones.
You can make a mapping, you can get a good measurement in map model, but to get a very good geo ref you need drones with RTK and base station.
You cannot play with an AIR 3 or Mavic 3 Pro for example and expect a good geo ref, is an illusion, as errors will be pretty high, starting from take-over point.
Cheers.

Geo_Drone Posted at 5-20 03:51
Geo ref. will suck on this type of drones.
You can make a mapping, you can get a good measurement in map model, but to get a very good geo ref you need drones with RTK and base station.
You cannot play with an AIR 3 or Mavic 3 Pro for example and expect a good geo ref, is an illusion, as errors will be pretty high, starting from take-over point.

At least we agree on that.

For the 2cm of accuracy, can you confirm that you are talking about absolute measurement on a 3D model or a Raster?
For example if you measure a road that is 6.00m wide, can you expect to have a reading between 5.98m and 6.02m?

If you could give a try it would be really  cool. Especially since you often fly close to sea level which makes the  data even easier to verify.

Before and after calibrating the compass

LV_Forestry Posted at 5-20 04:10
At least we agree on that.

For the 2cm of accuracy, can you confirm that you are talking about absolute measurement on a 3D model or a Raster?

We have made for an oil company mapping at 50 km of pipes using Autel Evo 2 PRO and GCP's, without RTK. We also had Enterprise RTK but was not needed...
The number of drones used was 12, all with automated missions, just returning to base for changing batteries.
At generated map in Pix4D we had an error of 2 to max 3 cm after GCP corrections applied. GCP were applied on ground both left-right and fwd-bk each 100 meters, with a control distance of 2 meters each.
Elevation of ground was also corrected using 1 GCP at elevations that were more than 10 meters, using double pass at 85 degrees gimbal and 90 degrees gimbal.

Was a 2 month job including the Pix4D processing...
Now, with Air 3 I have tried for fun only to make a 3D photogram.model, using 2 GCP, horizontal and vertical, with waypoints and overlapping around 70%. After correction I had at one pension a measurement error of 8 cm to 10 cm, because overlapping was not enough and we have not used second pass on outer circle, as 3D models needs for precision...So I believe can be improved with slower speed for 85% overlap and a second pass with inclined gimbal at 45 degrees...
Was just for curiosity, to see if is able...it is, but the Geo coordinates have errors that are far from reality because GPS is not as precise as an Evo 2.

Another example:
Evo 2 Pro is able to replicate a mission with an error of 1 meter from designated path....
Air 3 had 5 attempts to replicate waypoints saved...all time errors were more than 4 meters, at one point we had almost 10 meters...all experiments made in same area with more than 20 sats strong signal.
Is all about how precise the GPS and computing is....For mapping, even a standard Autel E2P have one processing unit that is separate, only for autonomous flight, like M3 Ent have.

Hope to help you this infos.
Cheers.