Map alignment problem after using Propeller Aeropoints (PPK)

WebODM
#1

I’m trying to get to a workflow to make use of WebODM along with Propeller Aeropoints. Totally new to the whole GCP / PPK thing but willing to learn!

I was able to get around the EPSG not being available by switching to a Proj string (see other thread on that).

I then took the data from the CSV file generated from the GCP targets out of the Propeller system. Loaded the data into the GCP editor, tagged about 6 photos from each set, and generated the output file.

I then processed the map using fastortho and without loading the GCP, followed by another one loading the GCP.

I then compared the two by seeing how well they lined up with the underlying Google Earth image.

After applying the GCP data, some parts of the map are better and some are worse. Better and worse in this case relate to how closely it matches Google Earth since I really don’t have much else to go by.
Question 1 - is that a valid test?

Now…there are definitely things I could do better. I had limited time as I was fitting in between rain showers. So…

All the targets were quite close together - I didn’t have time to move around a lot and I didn’t want to put them where people would be walking around/driving etc. So they were closely clustered around the middle of the ortho area.

The targets were only out there for a little over 10 minutes - 15 max. 10 minutes is the minimum but I’m sure leaving them longer would improve things.

At the moment I can’t say for sure that using the GCPs is improving the situation. Maybe it will if I spread them out more but I’m more worried that I am doing something wrong with the setup of the GCP - e.g. using the the wrong codes or measurements.
I’ve uploaded the Propeller output and GCP file I created in the hope that someone smarter than me (most people frankly) might feel inclined to take a look and point out anything dumb I’m doing.

Thanks in advance!

Note - I had to create them as PDF to upload here.
gcp_list_wyckoff v2.txt - Notepad.pdf (22.2 KB)
Wyckoff, NJ.pdf (24.8 KB)

1 Like
#2

Tag all the photos in which each point is seen.

Perform the comparison without the fastortho.

I have an unpopular opinion about Google Earth images. In my opinion they are very well georeferenced, much better than most believe. Still, your survey should be better. So it’s not a good comparison to know how well your survey is. In any case, how do you perform the comparison specifically?

The comparison is only valid in the area enclosed by the control points.

From what you see in the attachments, the coordinates of the control points are given in one 3D system or two (one horizontal and one vertical) systems:

  • EPSG:6319: Geographic 3D, Latitude and Longitude in decimal degrees and ellipsoidal height in meters. Geographic 2D equivalent is EPSG:6318.
  • EPSG:6527 + EPSG:5703: Projected 2D bounded with Vertical CRS. Northing and Easting in US survey feet and Orthometric height in meters.

In your case, I would use EPSG:6318. The result will be projected to a UTM zone framed in NAD83(2011). It may have a larger map distortion than the projection for New Jersey, but it would spare the software the interpretation of esoteric units of measure.

You are welcome!

1 Like
#3

Thank you so much for the response. My apologies for not responding sooner - unfortunately all the emails from this forum tend to end up in the spam folder.

I’ll give the suggestion of EPSG:6318 a go, and maybe I’ll try again spreading the GCPs further apart to see if that helps.

1 Like