We have noticed these inconsistent elevation values with DJI Phantom point clouds. For other UAS platforms (higher-end models), the shift tends to be consistent. I hope you find a solution but for ground control to be true ground control you would have needed to place targets out during the flight.
--
Jarlath O’Neil-Dunne
University of Vermont | USDA Forest Service R&D
On 8/4/18, 10:45 AM, "Northeast Arc Users Group on behalf of [log in to unmask]" <[log in to unmask] on behalf of [log in to unmask]> wrote:
This is primarily a visual exercise (viewsheds) so error in the transformation of elevations from drone measure to real world measure is acceptable and expected. Just want to get to roughly the right z values as grad student is working on thesis (and perhaps publications). The points are pretty much spot on in xy space. All the data were collected on the same day; over the course of approx. 7 hours.
Area of interest is +- 300,000 square meters small enough that we are not concerned with curvature of the earth surface. Also found out that the origin is different than what I was given before – so take off location and altitude of the drone is:
38.8 m Above Sea Level
33 deg 57' 19.93" N
35 deg 35' 49.15" E
At this location, using Google Earth for a ground control point (thanks for the suggestion Dave Tewksbury), the elevation is +- 28 meters.
The point cloud elevation at this location in the .las file is -90.093
As others have suggested, should be simple math (!), but just taking the home point elevation and trying to adjust the point cloud elevation based on that does not work.
Home point = 38.8 meters Above Sea Level
Point cloud elevation at Home point = -90.09 meters
Real Elevation (approximate from Google Earth but the only thing I have to go on) = 28 meters
Add home point to point cloud elevation gives -51.293 meters nowhere near 28 meters not to mention still negative.
Tried following as well – took several locations and found their real elevations (again approximate from GE) to use as GCPs
(-90.093) x (Y) = 28
Solving for Y
28/-90.093 = -0.3107
At a second location in the study area
33 deg 57’ 17.09” N
35 deg 35’ 43.24 E
At this location, using Google Earth for a ground control point, the elevation is +- 32 meters.
The elevation at this location in the .las file is -73.592
Applying the same math
(-73.592) x (Y) = 32
Solving for Y
32/-73.592 = -0.4348
This seemed promising – maybe the conversion is somewhere around -0.5 give or take
But then:
At a third location in the study area
33 deg 57’ 21.16” N
35 deg 36’ 7.81” E
At this location, using Google Earth for a ground control point, the elevation is +- 77 meters.
The elevation at this location in the las file is -12.992
Applying the same math
(-12.992) x (Y) = 77
Solving for Y
77/-12.992 = -5.927
Way different than the other two
I tried with four or five more GCPs and each time came up with very different Y values, so there is not a consistent one I can use across the entire point cloud.
Then I tried opening in Cloud Compare (thanks Neil Olson for suggestion – I had forgotten about this app). It does not open correctly but rather in a long single strip. Makes me think some important value is being transposed? Not sure…since I cannot “see” it, I cannot bring in the text file of GCPs to do point matching.
Conclusions so far:
- I am not using my GCP elevations correctly (bad math)
- Maybe over the course of the day, the drone’s home point was reset (I do not know) and Photoscan is able to adjust for this and stitch the data together to produce an .las file that is correct relative to itself. If so, maybe there is metadata in Photoscan that will say what the adjustment was? Sadly I cannot check as I do not have access to the software, only the output .las file
- Something about the coordinate system? XY are in decimal degrees, Z in meters. Not finding a tool in ArcGIS to project the .las file….will explore further.
Biggest conclusion:
NEARC folks are the best (although I already knew this). Thanks so much to everyone for the help.
------------------------------------------------------------------------- This list (NEARC-L) is an unmoderated discussion list for all NEARC Users.
If you no longer wish to receive e-mail from this list, you can remove yourself by going to http://s.uconn.edu/nearcsubscribe.
------------------------------------------------------------------------- This list (NEARC-L) is an unmoderated discussion list for all NEARC Users.
If you no longer wish to receive e-mail from this list, you can remove yourself by going to http://s.uconn.edu/nearcsubscribe.
