Proving UAV model accuracy

[James]

I have a question regarding aerial surveys. A local Registered Surveyor asked me to demonstrate/prove the accuracy of the drone generated model.
I have used Pix4d and Photoscan briefly in the past.
What are some ways to demonstrate the accuracy?

• include some GCP’s that are not part of the processing and check their location/height in the model?
• leave a staff flat on the ground at a fixed length and measure its length in the model? Maybe two at a 90 degree angle?
  Thanks,
  James

[Simon_Allen]

I address this problem regularly.

1. Collect a second set of ground verification points that are not used in the adjustment of your photogrammetry.
   Import these points and your image into a GIS and compare the verification points to their corresponding points on the image. Note . These points should be both inside and outside of the bounding box of your GCP’s to verify the whole image.
2. Your idea with a staff is a good idea, I use a laser range finder between marks further apart otherwise the pixel resolution is a significant part of the uncertainty.

Big Note. If you collect the GCP’s and GVP’s in realtime ensure that you are using exactly the same coordinates for the reference station in the field and during post processing. If you use base averaging prior to collecting your ground points and your flight and you use the Rinex header (from the rinex files downloaded from the reach) position in post processing you will NOT be using the same reference position. In my view this is a failure in the current reach software / workflow. The only way to ensure you are using the same base station coordinates is to write them down in the field and type them into the station position in RTKpost.

[James]

Awesome, thank Simon.
I found two articles by M. R James to be really useful.
One is:
3‐D uncertainty‐based topographic change detection with structure‐from‐motion photogrammetry: precision maps for ground control and directly georeferenced …

Kind regards,
James