Focus Encoder calibration for tracked cameras

 

...so how is this actually supposed to be done in 2024.2.0?

The tutorials simply tell me to use the basic calibrator, with no details how. The automated calibrator in Broadcast DE has no provision to calibrate focus at all, the values always remain at 0.

The forum shows me this:

https://my.aximmetry.com/post/3383-ue-tracked-camera-focus-distance-with-pthttps://my.aximmetry.com/post/3383-ue-tracked-camera-focus-distance-with-pt

but doing it this way completely screws up the calibration file drawn up in the automated calibrator. The whole floor gets distorted like a fisheye lens...

Eventually I will have to do this in software within the Aximmetry Flow itself anyway (see here  https://my.aximmetry.com/post/4045-bmd-lens-tracking-json-input-into-aximme ) but I can't even get it to work conventionally with a mechanical EZtrack Encoder yet. What am I missing?




   Stefan Reck

 
Profile Image
TwentyStudios
  -  
This is a big problem with the Camera Calibrator. The lack of focus calibration as well as the inconsistent, often inaccurate results makes me feel that Aximmetry needs to give their immediate attention and love to this product. As it is now it’s semi-broken. 
 
Profile Image
Stefan Reck
  -  

Right, so what's your workflow for calibrating tracked cameras? I found it really hard to get usable results in Basic Calibrator; the automated version at least puts out 6 somewhat working calibration points, albeit with a little too much "warbling" for my taste.

How many calibration points do I need to improve on this, and what actually happens behind the "radial distortion" values? And how would I get focus values into an existing calibration file without breaking it? The other post mentioned creating "substeps" - how does that work?



 
Profile Image
TwentyStudios
  -  

Use the regular, not Basic version of the calibrator. We use Sony cameras with auto focus and no ability to send the focus data into Aximmetry, so I’ve never delved into the workflow for focus calibration. From what I understand, you completely the calibration in the regular calibrator and then import the profile into the basic calibrator to add focus calibration points. If it breaks when doing this it might be a bug? 

 
Profile Image
Stefan Reck
  -  

yes, the camera calibrator in Broadcast really has some issues... When setting up multiple files for cameras with different lenses they seem to interfere with each other, so I have to do one camera, export the file, clear the calibrator and do the next one. Also the calibrator likes to put absurdly high values into the delta transformation which makes it pretty much impossible to calibrate our CN-E 70-200 lens. After calibration the Zoom works, but the floor appears to be shifted down by at least two meters. All the while these values should actually be zero, https://my.aximmetry.com/post/1522-delta-camhead-transtransbtransc


as we have a well calibrated 6DOF EZtrack system that already accounts for the translation between the camera sensors and the trackers.

I´ll probably have to resort to manually editing the XML files. Is there a way to import them into aximmetry without going through the calibrator? Are the file paths specified here still valid? https://my.aximmetry.com/post/1903-calibration-preset-does-not-show-up-as-a

 
Profile Image
TwentyStudios
  -  

If you already have it calibrated through EZTrack you don’t need to use the Aximmetry calibrator at all! If you’re calibrating a PTZ camera you still need to set the floor height manually and precisely. 

 
Profile Image
Stefan Reck
  -  

EZtrack is calibrated in itself and returns plausible values in all six degrees from origin, but does not give me a lens calibration file. So I would need to use the Aximmetry calibrator for that. With a CN-E 18-80 it does give me decent results, though it still puts some correction values into the Delta transform. What's weird is that these actually seem to be needed; resetting them manually makes the result noticeably worse... Maybe the calibrator sets up a lot of things here that cancel each other out?

 
Profile Image
Stefan Reck
  -  

So we checked our tracking system once again, tightened up the back focus, played with the lighting and finally got the CN-E 70-200 to calibrate as well. Well, sort of... We had to use the small calibrator board due to the limited space.

The lens calibration values look reasonable, apart from a few weird radial distortion values. The delta transform values however are complete nonsense. Using the file as it comes out of the calibrator the tracking is all over the place. The only way to somehow get it to work is copying the delta transform values from the other camera (CN-E 18-80) that has an otherwise completely identical setup and manually overwriting them. Again, leaving them at zero like Aximmetry suggests does not really work either.

Even then the results are somewhat disappointing. I get that it is really hard to accurately track 6DOF with a 200mm lens over a relatively simple IR laser system, but even at 80mm the tracking warbles and jerks a lot more than on the other camera at the same focal length. Any tips for optimizing this?



 
Profile Image
Eifert@Aximmetry
  -  

Hi Stefan,

You should determine which values from EZTrack are calibrated. You can use the Camera Tracking module for this, as explained in this guide:
https://aximmetry.com/learn/virtual-production-workflow/preparation-of-the-production-environment-phase-i/tracking/how-to-set-up-tracking-systems-in-aximmetry/
However, note that this might not be fully reliable because the tracking system might not fully report its (raw) tracking data through the protocol.



Unfortunately, with high zoom lenses such as the 70-200mm, you will likely encounter issues with the Aximmetry Calibrator in its current form. We are working on improving the camera calibrator to better accommodate higher zooms. In the meantime, you could perform the Tracking Calibration with a smaller lens on the camera, and then switch to the 70-200mm lens for Lens Calibration.

Warmest regards,

 
Profile Image
Stefan Reck
  -  

So as an update here, we pretty much had to ditch the automated camera calibrator and go with the basic calibrator and all the hassle that comes with it. There are various issues with the automated calibrator that make it unusable in our case:

- The automatic lens calibration seems to be achieved with a lot of center shift and radial distortion to get the position on point even with the focal length slightly off. This creates a lot of warble in both dimensions when zooming in and out of a static shot which looks really bad even on the 18-80mm. This is especially true at the longer end of the focal lengths.

- The tracking calibration results are consistently wrong at least in the z-axis (third triple). It will report the tracker as being well behind the camera sensor plane (negative value) when it is in fact slightly in front of it.

- Anything above about 60mm is basically impossible to calibrate automatically. The center shift and radial distortion values go through the roof, and running the tracking calibration with the 70-200mm wide open creates absurdly high delta head transformation values.

 
Profile Image
Stefan Reck
  -  

In the end we decided to get in someone from EZtrack to calibrate our lenses and generally wring out the best performance from the Lighthouse system, and the results are surprisingly good even at relatively long focal lengths. He did this with a combination of EZTracks EZprofile and quite a lot of painstaking hand optimization. With that in mind I have to conclude that while Aximmetry is a solid infrastructure and engine environment, the lens calibrator is simply not up to scratch (yet...). EZprofile offers a lot more measurement options that give you feedback about where your geometry is still wrong and how well the tweaks you apply are working. It is by no means automatic and a lot of work, but the results are worlds apart from what you can do inside Aximmetry. The Aximmetry basic calibrator theoretically gives you manual access to all the parameters, but without proper measuring tools getting all of them them just right is pretty much impossible.


;