Geometrically Verified Matching

Hi,

Geometrically verified matching is said to “improve robustness”, but there is no info in the help about what issues it might help to address. 

Will it help to identify tie points that otherwise would be rejected? Or to reject tie points that would otherwise be accepted?

If I have a set of photos that is a little “thin”, and from which pix4d without this option creates several submodels, might this option help? 

Thanks,

Hamish

Was this replied to? I was wondering the same. What was the outcome and answer?

Dear Hamish, John,

It is rather the second option: spurious matches will be rejected. You can find a detailed explanation of the effect of this option at the bottom of the page here

https://support.pix4d.com/hc/en-us/articles/205433155-Menu-Process-Processing-Options-1-Initial-Processing-Matching#gsc.tab=0

This option is most useful when there are repetitive structures (think for example of the windows of a building, where many window corners look exactly similar). 

Best regards,

1 Like

Would this then be an ideal setting when generating a 3D model of a lattice style cell tower?

Hi Aerovision,

It is hard to give a standard template for reconstructing cell towers because it will strongly depend on the acquisition plan, image, resolution etc.

Usually we recommend to users to do their tests on the step one to see how their can improve the calibration. For example, testing with and without GVM, with and without Rematch. As it is difficult to provide a general template that could be used for any data set. Photogrammetry is little bit more complicated than that.

However we have done some tests on our side and for cell tower reconstruction, we recommend to disable the option “Multiscale”.
See below the difference:


As the option multiscale will search for additional 3D points at different scale, it often lead to more noisy point cloud.

Hope this will help.
Best

1 Like