Support Website Contact Support Blog

QR Accuracy Issues


Just completed processing a 3D model with 5 GCP’s. GCP’s were input in NAD83(2011), with NAVD88 vertical system, but input as arbitrary since it’s not offered in Pix4D. First question is, did I do this correct?

I reoptimized after manually matching images for each GCP until I saw that it snapped to the center of each of the visible GCP’s I had placed on the ground.

I thought everything seemed to check out just fine, except my QR seems to be showing otherwise on the geo/relative accuracy of the entire project, please see the QR attached:

Is there anything else I should have done differently? the GCP part of the QR is showing very low error in the three coordinate directions. I am just trying to see what I should do differently, if anything, and also in simple terms what I am looking at. Is what I am seeing accurate enough? Not at all?

Thanks in advance!

It is a bit difficult for us to assess how the project came together without the full quality report. The screenshot you provided won’t tell the full story. Can you upload it and someone will take a look at it?

100144983 Model_report.pdf (614.5 KB)

Sorry, here you go. All of the questions I had were in my last post. Thanks again!

Hi bkeane,
I looked at your quality report and it looks very good. The RMS error for your GCPs was only 0.012 m. At this point I would be happy with your results. However, in the future you may want to incorporate check points as these are the best way to assess the accuracy of your model. They are used to give you accuracy statistics fully independent on the processing. Check point are not used in the optimization only for the accuracy reporting.

Thanks so much for checking. So with this being said, even though I do not have checkpoints, does that mean that my project is at the very least relative accurate? And if I had checkpoints in there then I could confirm absolute accurate? Just trying to understand better.

Thanks in advance!

There is no indication that your project has accuracy problems and I would feel comfortable using the data. Many users are happy at this point. However, the best way to assess the quality is to use a checkpoint or two as it will independently verify your accuracy.

Got it thanks. So those red areas in the absolute geolocation variance and relative geolocation variance, what are those and are they anything to worry about? They were in the images that I uploaded on the initial post on this thread. Can you explain those as well?

Thanks again!

This is talking about the image locations. In non-RTK drones the GNSS positions can be off by quite a bit.
Typically in the Z axis. If you are using GCPs, which you are, then this is not much of a concern as it will tie the project to its correct position.