Altitude errors

Joined
May 13, 2018
Messages
139
Reaction score
39
Age
78
My P4P is giving me good AGL altitude data in the downlink, and the Lat/Long GPS results are spot on. But the GPS altitude recorded in the EXIF photo data is showing ~ 20 meters below sea level when I'm really ~ 140 meters above MSL. I can't figure out what calibration might do that, but I'd have thought that a bad GPS unit would be bad all the way around. What am I missing? Can the GPS Z-axis really be that far off as a normal function?
 
My P4P is giving me good AGL altitude data in the downlink, and the Lat/Long GPS results are spot on. But the GPS altitude recorded in the EXIF photo data is showing ~ 20 meters below sea level when I'm really ~ 140 meters above MSL. I can't figure out what calibration might do that, but I'd have thought that a bad GPS unit would be bad all the way around. What am I missing? Can the GPS Z-axis really be that far off as a normal function?
The problem is the way DJI calculates that "GPS Altitude" number in the Exif data.
Despite the label, it doesn't come from GPS at all.
It comes from barometer data, very poorly converted to a nominal height above sea level using a standard value for normal atmospheric pressure.
But atmospheric pressure varies all the time and this does not take that into account.
It's commonly out by 200 ft or more.

Recalibrating things won't make any difference.
For anyone confused about this, it has nothing to do with the altitude you see on your screen, which is always reasonably accurate.
 
The problem is the way DJI calculates that "GPS Altitude" number in the Exif data.
Despite the label, it doesn't come from GPS at all.
It comes from barometer data, very poorly converted to a nominal height above sea level using a standard value for normal atmospheric pressure.
But atmospheric pressure varies all the time and this does not take that into account.
It's commonly out by 200 ft or more.

Recalibrating things won't make any difference.
For anyone confused about this, it has nothing to do with the altitude you see on your screen, which is always reasonably accurate.
That's really important information if you're a drone mapper. I produced photogrammetry mapping files that had good relative accuracy and precision in the statistics, but it was waay far off on vertical location. In order to solve that problem cleanly, I ended up buying a bulk editing app and changing that entry in the EXIF data for all the images. That worked a treat, but it was sure 'nuff a big surprise and a huge PITA troubleshooting problem.

Thanks a ton for the insight Meta4. I'd heard that it could be off, but man that's a lot. ............ R
 
That's really important information if you're a drone mapper. I produced photogrammetry mapping files that had good relative accuracy and precision in the statistics, but it was waay far off on vertical location. In order to solve that problem cleanly, I ended up buying a bulk editing app and changing that entry in the EXIF data for all the images. That worked a treat, but it was sure 'nuff a big surprise and a huge PITA troubleshooting problem.

Thanks a ton for the insight Meta4. I'd heard that it could be off, but man that's a lot. ............ R


Here's an example of what you are talking about:
i-MqRGc68-M.jpg


For mapping, the solution is to use Ground Control Points.
i-BnNNCqD-X2.jpg
 
I don't know what SFM software you use, but we have encountered this issue with the Phantom many times. For us, when we had to fly multiple missions to cover large acreage, especially at lower altitudes (like 35m) we would have each flight get tagged in the exif with wildly varying altitudes - most recently we went from 35m to 117m in the EXIF. What this did was cause the SFM software to make a lot of blocks because of the way it processes the initial step. If you use Pix4D mapper you can load the photoset, then highlight the first photo line, right click the box in the line that is under altitude, and you can tell it the altitude you flew. Once this is done it will change all the other photos in the set to that altitude for processing. You aren't affecting the EXIF of the photo, just the information the software reads from it. It's like changing a column of information in Excel. After that, it will process much more seamlessly. I don't know if other SFM software has that capability, but Pix does. I'd recommend always using GCP anyway, or at least RTK/PPK. After this step no matter what SFM software you use you will snap the model to the correct points as in Meta4's example.
1576685034229.png
 
Last edited:
  • Like
Reactions: modbuilder
The p4p also records the barometric altitude relative to the take off point, so won’t be (very roughly!) consistent anyway unless all take off points were exactly the same elevation. I use ground control points for mapping to calibrate if I need the most accurate dataset plus record location of take off locations.
 
I also use Pix4d but had not yet learned of the editing feature of the GCP manager. Would have saved me some $$. But I got the job done that triggered the post, and since then have done another mapping job, also with Pix4d and now using ArcGIS Mapper to overlay as-built orthomosaic and contours on the architect's plan. Kinda cool.

Regarding the altitude to be used in mapping engine, that entry needs to include the difference between the geoid and the ellipsoid heights at that location anyway. So it's not going to happen that any of the EXIF altitudes in the photos can be used as-is anyway. LIke AxelZ, I also shoot a rover fix at the takeoff point and use that for later finding the altitude to be entered in the GCP table.

Having the right altitude makes Pix4d a lot more accurate at grouping the photos to be marked as GCP's. i guess I was just not expecting the EXIF absolute altitude entry in the P4P's meta data to be an absolute fiction. The baro relative altitude is useful, but otherwise the altitudes are useless. Maybe the P4P RTK edition has this sorted out, but I have no knowledge of that bird.
 
Last edited:
I also use Pix4d but had not yet learned of the editing feature of the GCP manager. Would have saved me some $$. But I got the job done that triggered the post, and since then have done another mapping job, also with Pix4d and now using ArcGIS Mapper to overlay as-built orthomosaic and contours on the architect's plan. Kinda cool.

Regarding the altitude to be used in mapping engine, that entry needs to include the difference between the geoid and the ellipsoid heights at that location anyway. So it's not going to happen that any of the EXIF altitudes in the photos can be used as-is anyway. LIke AxelZ, I also shoot a rover fix at the takeoff point and use that for later finding the altitude to be entered in the GCP table.

Having the right altitude makes Pix4d a lot more accurate at grouping the photos to be marked as GCP's. i guess I was just not expecting the EXIF absolute altitude entry in the P4P's meta data to be an absolute fiction. The baro relative altitude is useful, but otherwise the altitudes are useless. Maybe the P4P RTK edition has this sorted out, but I have no knowledge of that bird.

You could try contacting DJI. They fixed the problem in the Mavic 2 firmware (Firmware v01.00.0510 EXIF altitude changes) but it doesn't look like they addressed it with the P4P, which is unfortunate since it probably sees more use for these kinds of application.
 
That's a good suggestion. However, I do as little with DJI as possible, simply because of the GeoFencing. I fundamentally oppose the philosophy and only use their equipment because there's been no affordable good alternative until the EVO 2 release. I'm OK continuing to run aerial maps without a good absolute altitude in the metadata now that I know how to work around it.

In actuality, my mapping jobs usually have to work with land surveyor coordinate system data and elevations based on monuments. The only way to be sure I'm producing comparable data is to shoot a monument and use that a reference for my precision GPS registrations.

Thanks for all the feedback. It's helped. Regards............... Bob R.

(SAR104: That's pretty close to the tag line used by Walter Brennan in The Guns of Will Sonnet : "No Brag; Just Fact". But I suspect it's actually from Dragnet. Now I've given away just how much of a geezer I really am" )
 
Last edited:

Members online

Forum statistics

Threads
143,066
Messages
1,467,356
Members
104,934
Latest member
jody.paugh@fullerandsons.