Which Optimize Camera Alignment Parameters should we use?

[DayGeckoArt]

The manual says what the different checkboxes mean on the Optimize Camera Alignment dialogue, but doesn't explain when or why you would use each one. The last 3 are disabled by default, but I've seen user made tutorial videos where they check all but without an explanation. I usually check all except the Advanced boxes, when I optimize alignment after adjusting my GCP markers. Is this correct? Or should I be using the default settings? Does it depend on the type of camera?

[dpitman]

I have not seen a good, not too technical, explanation about which parameters are relevant for which camera types. And a workflow strategy based upon the available settings in Metashape.
 
There is clearly an opening for someone that knows the subject well enough to teach it to others to present a white paper or video. Or, perhaps the material already exists and someone could share it?

[jenkinsm]

It exists:

https://pubs.er.usgs.gov/publication/ofr20211039

[dpitman]

That report is very detailed and useful for sure.  But it doesn't really go into the camera coefficients.  Mostly just suggesting:

“Optimize Cameras” with one of the following options:
• default coefficients selected [f, k1, k2, k3, cx,
cy, p1, p2];
• all coefficients selected [f, k1, k2, k3, k4, cx, cy, p1,
p2, b1, b2] and “Fit additional corrections”; or
• some other combination of coefficients that fits the
data and camera.

I suspect with the cameras on consumer drones, with maybe the exception of the P1, using the default coefficients are probably best if you don't know any better.  Thoughts?

I found it refreshing to see that the report authors included the following commentary on MS's use of the term "cameras" which agrees with what I thought when I started using the application.  The application is complex enough without adding unnecessary confusion.  Not sure why Agisoft doesn't correct this which was probably just the result of translation.  Perhaps they think it is too late and just live with it.

Note that the term “camera” in Metashape is used to describe a specific image at a specific time and location, and the term is also
the tool used to take photographs. Metashape uses “image”, “photograph”, and “camera” interchangeably. This creates an unwanted complexity in the software;

[Corensia]

dpitman,

I thought the same about the term 'camera' until I started using thermal image sets. There is a distinction between Metashape's use of image and camera. For example, two images taken from the same spot (one RGB and one Thermal) can be configured as one 'camera' with two images. So while the term camera seems interchangeable with image, it's actually different since 'camera' represents the exact time and conditions when the image was taken and several images can be associated with one 'camera'. I agree that It definitely can be confusing as an English speaker though.

[dpitman]

Hi Corensia,

Unfortunately, the word "camera" already had an English definition before the existence of Agisoft (and even photogrammetry for that matter).  And, I believe Agisoft was using the term (incorrectly IMO) before Photoscan started working with thermal imagery.

The fact is, that MS processes "images" (or "photos") and not cameras.  No matter what they call them.  It's not a problem for me now. It was just confusing when getting started.  My comment about the USGS paper was simply that those folks, probably a few with doctorate degrees in geo-science, agreed with this enough to mention it.  As to why the Agisoft folks leave the error in place is unclear.