I have a client who specifically asked for videos with video captions for AI use, who I found on Upwork. I turned the function on on DJI 4, and the altitude has been giving me a headache.


After reseaching for a day I found that the altitude reading on video captions of Mavic 2 Pro is ASL which suggests approximately how high the drone is from the sea level. This is different from the altitude reading on my controller which says "height", it is obviously the height from the ground (and more precisely the current height from the level of take-off point). My client has rejected my test videos as they need the altitude that is the height from the ground, not how high from the sea.

This is dissapointing that the video captions will not record the height, what we see on my controller which would suffice their requirements. And for that I might not be able to do the job. I have suggested my client that I will take the videos from the ground so the difference from the ground to the flying level would be relatively suggested, as I tested several times and found that the difference in readings between the ground level and the flying level is relatively accurate no matter what the altitude reading on the ground is (it is never 0, because I am not launching from the sea surface. It is typically between 10 to 100m), just like when you rely on the controller's display of "height"

For the limited display on the video captions on Mavic 2 Pro, I may not be able to conduct this job. All I need is Mavic 2 Pro's SRT file can display what I see as its height on my controller. Is there any way around or fix for this?

I have uploaded the screenshot that I took of a video captured with my Mavic 2 Pro. It says "altitude: 170.382996", and I was flying about 115m. So the ground is about 55m above sea. This altitude is useless to my purpose.

Image found on https://mavicpilots.com/threads/altitude-reading-for-my-video.128441/#post-1627379

I found that the altitude reading on video captions of Mavic 2 Pro is ASL which suggests approximately how high the drone is from the sea level.
It isn'tthe height above sea level at all.
There's no way that the drone can tell it's height relative to sea level.

Labroides Posted at 3-10 03:38
I found that the altitude reading on video captions of Mavic 2 Pro is ASL which suggests approximately how high the drone is from the sea level.
It isn'tthe height above sea level at all.
There's no way that the drone can tell it's height relative to sea level.

Then what's the value supposed to be?

Labroides Posted at 3-10 03:38
I found that the altitude reading on video captions of Mavic 2 Pro is ASL which suggests approximately how high the drone is from the sea level.
It isn'tthe height above sea level at all.
There's no way that the drone can tell it's height relative to sea level.

EXIF from mavic 2 :

Hi, Yostan0526. Thank you for reaching out and we're sorry for the inconvenience. Can you please upload the screenshot to a shareable link like Google Drive or DropBox for us to further check? The link above is not accessible on our end. Thank you.

The altitude reading of my video captions is 170.382996. I was flying about 115m above the ground (as my controller indicated). The ground elevation is about 60m above the sea.

Excuse my ...ty image. I took an image about 29m above the ground (my controller suggested 29m) and its metadata says its altitude is 84m. I believe the elevation of the take off point (I just flew right above the take off point) has been added to the "height" as displayed in my controller 29m.

I have put the link for this image's DNG file
https://drive.google.com/file/d/1oNRcj903c3gburl6xW0T54MbB2n3n7--/view?usp=sharing

Yostan0526 Posted at 3-11 02:17
[view_image]

Excuse my ...ty image. I took an image about 29m above the ground (my controller suggested 29m) and its metadata says its altitude is 84m. I believe the elevation of the take off point (I just flew right above the take off point) has been added to the "height" as displayed in my controller 29m.

I took an image about 29m above the ground (my controller suggested 29m) and its metadata says its altitude is 84m.
Perhaps you meant 39 metres rather than 29.
The metadata for your image contains two different heights. ... See the screenshot below.
Absolute Alt = +84m
Relative Alt = +39.7m

I believe the elevation of the take off point (I just flew right above the take off point) has been added to the "height" as displayed in my controller 29m.
The two different heights have nothing to do with each other.
The drone has no way to tell what the elevation of the launch point was

The relative height comes from the barometric sensor, showing the height relative to the launch point.
The absolute alt is a rough attempt at calculating height above sea level.
But because DJI failed to account for the normal fluctualtions in air pressure and assumed standard air pressure, the result is garbage and varies wildly depending on the weather.
The error can be +/- 100 metres or more.

Labroides Posted at 3-11 05:36
I took an image about 29m above the ground (my controller suggested 29m) and its metadata says its altitude is 84m.
Perhaps you meant 39 metres rather than 29.
The metadata for your image contains two different heights. ... See the screenshot below.

Awesome it did have the relative altitude! Perhaps it was 39m above. What did you use to get the data from my image?

Yostan0526 Posted at 3-11 04:09
I have put the link for this image's DNG file
https://drive.google.com/file/d/1oNRcj903c3gburl6xW0T54MbB2n3n7--/view?usp=sharing

So far for me everything is fine and accurate. 84m absolute (WGS84), which gives 49m (NZVD16 (sea level)).Despite what has been written, yes the drone can know its starting position in relation to sea level. I never asked myself the question of whether the Mavic2 could do this but it is possible for any GNSS receiver which integrates the EGM96. This is a 2Mb file which indicates the difference between the GNSS altitude (WGS84) and the altitude above mean sea level. For the whole planet!.


The park on the corner of Parnell Road and Ruskin Street is 40m above sea level. If you put your drone on the ground and take a photo, it should be around 74.5m (absolute).

LV_Forestry Posted at 3-11 09:29
So far for me everything is fine and accurate. 84m absolute (WGS84), which gives 49m (NZVD16 (sea level)).Despite what has been written, yes the drone can know its starting position in relation to sea level. I never asked myself the question of whether the Mavic2 could do this but it is possible for any GNSS receiver which integrates the EGM96. This is a 2Mb file which indicates the difference between the GNSS altitude (WGS84) and the altitude above mean sea level. For the whole planet!.

So far for me everything is fine and accurate. 84m absolute (WGS84), which gives 49m (NZVD16 (sea level)).  Despite what has been written, yes the drone can know its starting position in relation to sea level.

If that was true, the Absolute Altitude recorded from the same position, would remain the same for images shot at any time/date.

But Absolute Altitude is not constant.
It's all over the place because in DJI consumer drones it uses the barometric sensor, not GPS:

Yostan0526 Posted at 3-11 08:45
Awesome it did have the relative altitude! Perhaps it was 39m above. What did you use to get the data from my image?

Awesome it did have the relative altitude! Perhaps it was 39m above.
Not awesome ... your video subtitles show DJI's completely useless "Absolute Altitude"
It shows 43 metres at takeoff.
If it was showing the relative altitude, it would show zero metres.
Shoot on a day when the air pressure is different and the Absolute Altitude shown in the subtitles will be different.

What did you use to get the data from my image?
Most simple metadata readers only show Exif data which only has the Absolute Altitude.
You need a more sophisticated metadata viewer to show the full metadata including XMP which contains both Altitude fields.
That one is called Picture Information Extractor.
It's good for still images, but won't help for video files.

Labroides Posted at 3-11 17:00
Awesome it did have the relative altitude! Perhaps it was 39m above.
Not awesome ... your video subtitles show DJI's completely useless "Absolute Altitude"
It shows 43 metres at takeoff.

Yeah although I do not believe the absolute altitude is inaccurate, the problem of displaying the relative altitude is yet to be solved. It would be the easiest if it could just write in AGL (since it actually stores the values)

I am researching at the moment if I can rewrite the altitude on SRT files with some third party generated data files that show relative altitudes, or probably I could just do the calculation of (absolute altitude on the flying level - the altitude at the ground level) on Excel and replace all the altitude data on the SRT files (and compare the values with relative altitudes obtained using some third party generated data viewer) That will probably provide usable data (I do not think my client will penalise me for putting data that are apparently ok but technically inaccurate)

Yostan0526 Posted at 3-11 17:51
Yeah although I do not believe the absolute altitude is inaccurate, the problem of displaying the relative altitude is yet to be solved. It would be the easiest if it could just write in AGL (since it actually stores the values)

I am researching at the moment if I can rewrite the altitude on SRT files with some third party generated data files that show relative altitudes, or probably I could just do the calculation of (absolute altitude on the flying level - the altitude at the ground level) on Excel and replace all the altitude data on the SRT files (and compare the values with relative altitudes obtained using some third party generated data viewer) That will probably provide usable data (I do not think my client will penalise me for putting data that are apparently ok but technically inaccurate)

Yeah although I do not believe the absolute altitude is inaccurate
How would you explain the data from the channel marker photos above?
I've got plenty more showing the same variations.
DJI's absolute Altitude is absolutely useless.
It changes with the weather and this is easily demonstrated.

Labroides Posted at 3-11 18:00
Yeah although I do not believe the absolute altitude is inaccurate
How would you explain the data from the channel marker photos above?
I've got plenty more showing the same variations.

I cannot answer about your images, but as far as about the images I took, the altitude readings seem to be correlated to the absolute altitude, and I can still get close to the "height" displayed on my controller by looking at the difference of the altitude values taken at the ground level and taken at the flying level.

I can only imagine that your images indicate different absolute altitude values because: 1. the IMU was not calibrated on either/both, 2. taking the footage several months apart created different conditions for your drone's barometer of air pressure (perhaps seasonal changes in air pressure), and/or 3. the water under the drone gave it the wrong interpretation of its altitude. I believe my cases are affected by none of these especially as I have been calibrating IMU since I noticed this altitude issue. Before that I noticed that the altitude showed -70m

Yostan0526 Posted at 3-11 18:17
I cannot answer about your images, but as far as about the images I took, the altitude readings seem to be correlated to the absolute altitude, and I can still get close to the "height" displayed on my controller by looking at the difference of the altitude values taken at the ground level and taken at the flying level.

I can only imagine that your images indicate different absolute altitude values because: 1. the IMU was not calibrated on either/both, 2. taking the footage several months apart created different conditions for your drone's barometer of air pressure (perhaps seasonal changes in air pressure), and/or 3. the water under the drone gave it the wrong interpretation of its altitude. I believe my cases are affected by none of these especially as I have been calibrating IMU since I noticed this altitude issue. Before that I noticed that the altitude showed -70m

It's because DJI's formula to calculate Absolute Altitude is flawed.
They assumed STP - standard temperature and pressure, which is fine in a laboratory setting, but useless outdoors in the real world.
It does not alow for normal changes in air pressure due to weather.

Your drone's absolute altitude will swing up and down with the daily fluctuations in air pressure.
Compare the absolute altitude for shots at known identical actual altitude, taken at times of high and low atmospheric pressure.
That's what I illustrated with the channel markers.
I've been working on this for a few years to work out why DJI's absolute altitude swings all over the place.

Yostan0526 Posted at 3-11 18:17
I cannot answer about your images, but as far as about the images I took, the altitude readings seem to be correlated to the absolute altitude, and I can still get close to the "height" displayed on my controller by looking at the difference of the altitude values taken at the ground level and taken at the flying level.

I can only imagine that your images indicate different absolute altitude values because: 1. the IMU was not calibrated on either/both, 2. taking the footage several months apart created different conditions for your drone's barometer of air pressure (perhaps seasonal changes in air pressure), and/or 3. the water under the drone gave it the wrong interpretation of its altitude. I believe my cases are affected by none of these especially as I have been calibrating IMU since I noticed this altitude issue. Before that I noticed that the altitude showed -70m

I believe my cases are affected by none of these especially as I have been calibrating IMU since I noticed this altitude issue. Before that I noticed that the altitude showed -70m

Think again ....
If calibrating the IMU "fixed" an error of 70 metres or so in your absolute altitude, you would also have been getting the same error in your relative altitude.
What you observed was due to weather conditions.

And if you've recently recalibrated the IMU, doing it again probably won't fix the 6 metre drift in your recorded flight data.

Labroides Posted at 3-11 16:04
So far for me everything is fine and accurate. 84m absolute (WGS84), which gives 49m (NZVD16 (sea level)).  Despite what has been written, yes the drone can know its starting position in relation to sea level.

If that was true, the Absolute Altitude recorded from the same position, would remain the same for images shot at any time/date.
[Image]

I have never encountered this problem with the Mavic 2. I would like to believe that DJI has deliberately handicapped its consumer drones on this point by creating this GPS + Baro calculation, for post-phantom 4 models. Maybe !

LV_Forestry Posted at 3-11 20:42
I have never encountered this problem with the Mavic 2. I would like to believe that DJI has deliberately handicapped its consumer drones on this point by creating this GPS + Baro calculation, for post-phantom 4 models. Maybe !

Perhaps you've not looked closely at your image metadata.

It's got nothing to do with GPS.
It's the same on my Mavic 3 pro as it was on the Phantom 3 and Phantom 4 pro.

Labroides Posted at 3-11 20:59
Perhaps you've not looked closely at your image metadata.

It's got nothing to do with GPS.

No chance.  Since I have been using it for photogrammetry, not just me, we would have noticed it.

LV_Forestry Posted at 3-11 22:35
No chance.  Since I have been using it for photogrammetry, not just me, we would have noticed it.

You think photogrammetry uses absolute altitude?
You also think compass calibration has to do with your geographic location, don't you?

Photogrammetry with images from DJI consumer drones uses relative altitude.
I've done it for years with drones that exhibit the wildly inaccurate and swinging absolute altitude.

Labroides Posted at 3-11 22:40
You think photogrammetry uses absolute altitude?
You also think compass calibration has to do with your geographic location, don't you?

You think photogrammetry uses absolute altitude?


Agisoft Metashape screenshot :




You also think compass calibration has to do with your geographic location, don't you?

It's a fact, the Mavic 2 systematically requests a calibration of the compass after each trip greater than 50km. Phantom 4 too. I gladly invite you to come visit me whenever you want and see it.


Looking forward to seeing you in our hemisphere Owen.

Labroides Posted at 3-11 16:04
So far for me everything is fine and accurate. 84m absolute (WGS84), which gives 49m (NZVD16 (sea level)).  Despite what has been written, yes the drone can know its starting position in relation to sea level.

If that was true, the Absolute Altitude recorded from the same position, would remain the same for images shot at any time/date.

Once again for me everything is clear.

The images from the Mavic 2, that of the OP and mine, have an altitude expressed in WGS84 format. With the so-called WGS84 ellipsoid for reference.
In the EXIF it is mentioned "Above Sea Level" which is not the same as "Above Mean Sea Level", be careful the difference is important!

The altitude that is presented in the EXIF of your Mavic 3 Pro images is clearly mentioned GPSFusionAlt :


IMU and GPS Fusion for Inertial Navigation - MATLAB & Simulink (mathworks.com)

LV_Forestry Posted at 3-11 22:50
You think photogrammetry uses absolute altitude?

It's a fact, the Mavic 2 systematically requests a calibration of the compass after each trip greater than 50km.
Yes, it did ... and there was absolutely no physical reason for it to.
It was just DJI hanging on to the misinformation that had put in their manuals for years.

Phantom 4 too.
You'd think that I'd have noticed in the 6 years that I used Phantom 4 pros, travelling 5000 km east-west and 3200 km north-south ... but somehow I missed it

Sorry ... I'd rather deal with someone who deals with factual information..

I gave up on you ever understanding basic factual information when I tried to explain compass calibration myths.
It's like trying to talk facts to a menber of the Trump cult.

Labroides Posted at 3-11 23:14
It's a fact, the Mavic 2 systematically requests a calibration of the compass after each trip greater than 50km.
Yes, it did ... and there was absolutely no physical reason for it to.
It was just DJI hanging on to the misinformation that had put in their manuals for years.

The fact is also that I agree with you regarding the compass.
It's just that the Mavic 2, P4M, Ardupilot, Pixhawk that I use continually require this calibration. That's all.

I think it's not misinformation, it's a question of technology and algorithms.
For a project to guide an autonomous forestry machine I used a good old compass at the beginning. The problem is that from one end of the country to the other it did not show the same North, I am not going to redo the development on the magnetic declination...

Then I moved on to a solution calculated on the basis of the indications of the GNSS receiver (antenna only), Azimuth between coordinates A and Coordinates B when the wheels are straight. This solved the problem of magnetic North but with GNSS precision it also brought defects.

Then I coupled them to an IMU to discard the GPS faults, (the Mathlab link previously posted), this is very probably what DJI is doing in these new drones.

And finally I upgraded to Full GNSS with dual receiver. The distance between the two antennas being known. one is the base, the other the rover. With this the geographical heading is ultimate and almost instantaneous. And it works everywhere.

simpleRTK2B Heading - Basic Starter Kit (ZED-F9P) - ArduSimple

LV_Forestry Posted at 3-11 22:58
Once again for me everything is clear.

The images from the Mavic 2, that of the OP and mine, have an altitude expressed in WGS84 format. With the so-called WGS84 ellipsoid for reference.

Once again for me everything is clear.
Your reasoning isn't

The images from the Mavic 2, that of the OP and mine, have an altitude expressed in WGS84 format. With the so-called WGS84 ellipsoid for reference.
Maybe they are, but I can't see anywhere in DJI image metadata where this is stated.
I can't see anything to suggest that the data from the Mavic 3 isn't either.
What's your point?

In the EXIF it is mentioned "Above Sea Level" which is not the same as "Above Mean Sea Level", be careful the difference is important!
The Exif for many of my Phantom and Mavic 3 pro images says Below Sea level, when the images were clearly shot well above sea level..
I can't see your point.

The altitude that is presented in the EXIF of your Mavic 3 Pro images is clearly mentioned GPSFusionAlt :
So??
The XMP section for the metadata of the Mavic 3 pro has many fields (including Altitude Type) that weren't present in earlier models.
What's the relevance?

Don't bother responding till you can explain why the absolute altitude in the channel marker photos is so different.
That's something that actually matters.

LV_Forestry Posted at 3-11 23:32
The fact is also that I agree with you regarding the compass.
It's just that the Mavic 2, P4M, Ardupilot, Pixhawk that I use continually require this calibration. That's all.

Compass calibration has nothing to do with your local magnetic variation.
It's all about the magnetic fields that are part of your drone.

You are hijacking this thread as you've hijacked others before.
I've had this compass discussion with you before.
You just don't understand and I'm not going to go over it again.

Labroides Posted at 3-11 23:32
Once again for me everything is clear.
Your reasoning isn't

Once again, above sea level is different than above mean sea level. The region you live, south of Australia has a negative value referring to the ellipsoid. This explains why results are expressed as below sea level.

LV_Forestry Posted at 3-12 00:18
Once again, above sea level is different than above mean sea level. The region you live, south of Australia has a negative value referring to the ellipsoid. This explains why results are expressed as below sea level.

Perfect explanation ... except that many of my images  (from similar altitudes) have positive absolute altitudes and they show as Above Sea Level.
Those with negative absolute altitudes show as Above Sea Level.
The Absolute Altitude is all over the place and often nowhere near accurate.

So get back when you actually understand the issue.