Error in GCP correlation too large

PIX4Dmapper PIX4Dmapper Questions/Troubleshooting
1

I posted this in the Pix4d Desktop thread. re-posting here because I have received any responses and this topic fits better.

I’ve read the support article: https://support.pix4d.com/hc/en-us/articles/202969759-Can-the-Theoretical-Error-S-X-Y-Z-of-a-GCP-be-small-while-the-Error-to-GCP-Initial-Position-is-high-#gsc.tab=0

I have 23 GCPs in a project with 542 images, covering 90 acres with an average GSD of 3.67cm/1.44 in. The GCPs were surveyed with terrestrial RTK GPS instruments.  I’m getting very poor correlation between the surveyed GCP coordinates and the computed coordinates. The Theoretical error is low but the error to GCP initial position is high in the X and Y around 2-3x the GSD at almost all GCPs in either the X or Y directions. The theoretical and error to initial Z position are actually great - equal to or less than the GSD.

That’s with using all GCPS as 3d GCPs. If I mark any of them as check points the X, Y error relative to the initial position is more like 5x the GSD. 

What am I doing wrong? I’ve checked all of the GCP coordinates and they match what the surveyor delivered from his data collector. What else could be the problem? 

2

Hello,

The theoretical error that you are mentioning is giving a representation of the relative reconstruction accuracy and indicates how well GCPs were marked on the images. You mention that the theoretical error is small - this is good and it means that the GCPs are properly marked on the images and project is well reconstructed.

As for the Error to GCP Initial Position value. Keep in mind that the expected accuracy of the reconstruction is 1-2 GSD horizontally (X,Y coordinates) and 1-3 GSD vertically (Z coordinate). This means that the Error to GCP Initial Position is not exceeding the expected values.

Could you please submit a request to Support here and send us the quality report of the project in order to further investigate.

Regards,