Improve Accuracy by using DEM Height instead of EXIF

Discussion created by g.doerffelmysynergis-esridist Employee on Jul 16, 2017

Hi everybody,

I had recently processed a series of drone flights over densly closed forest. Hardly any ground visibility at all. Results were "sub-optimal".
The first thing I wanted to improve upon were very bad height-values stored in the EXIF of the images. I luckily have a DEM and was able to improve the results (especially the location/scaling) considerably.

The drone data was from a DJI 330 on a Phantom4 consumer drone. I learned that it actually only has a height value stored for the starting point and then the flight height as such. So the "individual" height values on each image were pretty off because the start-height was off. To improve, I followed this workflow:

  • Exported EXIF of all imges in a folder to csv using EXIF-Tools  (If I improve that, I'll use a python exif extractor)
  • Modify elevation/height in csv with height from Starting point PLUS Flight height
  • (Optionally also convert WGS84 X/Y into projected coordinate system including transformation
  • After loading images to Drone2Map replace the elevation by reading in the csv file 
  • Process ...

I have all this in python and would actually be willing to share as a toolbox - but would like to have some community-input first:

  • Is this the way to go?
  • Any suggestions what to include in addition?
  • Has anyone got an idea how "valuable" YAW/PITCH/ROLL values stored by a Phantom4 are?
    (and how to make use of them in Drone2Map?   Photoscan/AGIsoft seems to be able to read/use them)

So fire away ... :-)  Thanks