Hello,

the altitude in images exif is taggued as above sea level, but it's obviously not the case... I took pictures on the ground at a place aroud 800 m AMSL and the reported value is close to 0.
The GPS was very good. It look like the field is used to store the delta to ground or start point. I would much prefer having the real GPS altitude (without geoid correction : ellipsoidal height) instead of some "secret altitude recipe" done by DJI.
(altough publighing the recipie to cook the value as it is computed right now would be nice !)

Regards

No "secret recipt". The altitude DJI records is always AGL because it uses relative barometric pressure changes to do the calculation and because that's what's most important in piloting an aircraft..   GPS altitude is +/- 45 meters accuracy (reference ellipsoid relative).  Make's it tough to plan a landing :-)  

To my knowledge, the Spark GPS does not compute GPS altitude.  There IS an entry in the Mavic's DAT file for "Absolute Height", but if apparently requires an entry for Home Point Elevation - manual? - maybe that only applies to some of the commercial drones.  Probably not Spark

In any case, the geotaging fields in exif are all optional.  I doubt if many consumer product vendors can populate all 32 fields in the exif tag. You get what you get!

Griffith Posted at 2017-7-5 10:38
No "secret recipt". The altitude DJI records is always AGL because it uses relative barometric pressure changes to do the calculation and because that's what's most important in piloting an aircraft..   GPS altitude is +/- 45 meters accuracy (reference ellipsoid relative).  Make's it tough to plan a landing :-)  

To my knowledge, the Spark GPS does not compute GPS altitude.  There IS an entry in the Mavic's DAT file for "Absolute Height", but if apparently requires an entry for Home Point Elevation - manual? - maybe that only applies to some of the commercial drones.  Probably not Spark

Thanks for the recipe. From the picture I had, I believe this was it but was not sure.

As for the +/-45m in GNSS vertical accuracy, I think this is a bit exagerated (or given at 6 sigmas) . In full opensky environnent at the latitudes where I fly (not high latitudes and not latitudes where ionosphere and scintillations are often very perturbed), L1 code only, GNSS receivers provides usually better than 3 meters (you then have to add geoid separation to go from ellipsoidal to AMSL, and this offset can be 45 m, but it's pretty well tabulated for this level of accuracies)

GNSS altitude is indeed much more noisy than barometer altitude, but it does not drift (altough flight times of our small uav of course do not let us experience this and in the end it's not an issue)

Still the tag AMSL is misused in EXIFs

Matioupi Posted at 2017-7-5 11:18
Thanks for the recipe. From the picture I had, I believe this was it but was not sure.

As for the +/-45m in GNSS vertical accuracy, I think this is a bit exagerated (or given at 6 sigmas) . In full opensky environnent at the latitudes where I fly (not high latitudes and not latitudes where ionosphere and scintillations are often very perturbed), L1 code only, GNSS receivers provides usually better than 3 meters (you then have to add geoid separation to go from ellipsoidal to AMSL, and this offset can be 45 m, but it's pretty well tabulated for this level of accuracies)

As for the +/-45m in GNSS vertical accuracy, I think this is a bit exagerated (or given at 6 sigmas) .

GNSS IS actually much better.  I was quoting GPS alone (not GPS + GLONASS +?).  I typo'd the 45 :-) - should have been +/-35 meters (98% confidence). The tests, as I recall, were performed on cell phone GPS receivers - so not top quality - and done a few years ago. I think most new phone use multiple systems and are probably far more accurate now ( if enough sats are visible).

I don't know what DJI is actually using.