Reporting | X,Y,Z +/- Accuracy

Hi Pix4D Community,

I have been trialling Pix4D Mapper for a while and running tests using a DJI Mavic 3E comparing results for:

  2. RTK
  3. Zoom used/not used
  4. Flight speed maximum to capture data/image quality
  5. RTK + GCPs for PPK (not started yet)

On the Quality Report I can see the GSD accuracy e.g. 1.7cm but this only really tells me how many cm/pixel have been captured to my understanding.

What I am trying to identify from the report is what is the accuracy on the X,Y,Z planes in terms of actual accuracy e.g. +/- 1.5cm

I need to be able to understand which set/s of data in the report advise me of this so I have confidence in the output to be exported to CAD/AutoCAD for clients. They will have a request for surveying accuracy for say +/- 5cm and I need to confidently advise that we have achieved this accuracy.

Please can you advise where I read this figure from in the Quality Report, if indeed it is there?

At the moment an associate has advised I use the RMS Error [m] to obtain the +/- X,Y,Z but looking at the tests I have run these suggest that RTK is only a few parts of a cm more accurate than GPS which I just cannot believe is the case.

Many thanks

The best way to assess accuracy is to use checkpoints. This is the only independent method to understand how accurate your project is. The other metrics in the quality report do help in understanding how good your dataset is and how well it is processed. But you need checkpoints to reliably understand the accuracy of your project.

Thanks Mike, so just so I am crystal clear - for example:

  1. I am on a site to be surveyed
  2. 3-4 GCPs are placed in strategic areas to be ‘visible’ when processing Topo, Ortho, 3D Model
  3. Checkpoints - would they be the measurement I take using a rover at the centre of the GCP?

In effect what I am asking is, if I take a detailed measurement at the centre of the GCP then is THIS a Checkpoint (even if not processed in the data processing stage or even if it is)?

Sorry to ask, scant information of this level of detail that I can find. So I am trying to understand from a step by step approach.

Checkpoints will look just like a GCP. It will be measured and marked in your images exactly the same way. But instead of setting it as a GCP, you need to set it as a Checkpoint in the UI. If it is a Checkpoint, Mapper will not try to fit the model to this point. Rather, it will give you the computed error in the X, Y, and Z to what it thinks it should be.

This video will describe it in more detail: