How does sensor size affect TRACKING (in general)?

Hello. I don’t quite understand deeply what the relation of the real life camera sensor in relation the tracking with the virtual camera size is.

I’ve always worked with direct aspect ratio (i.e: 1920 x 1080). In the case of Arri Alexa sensor size to full 35 mm. And that’s about it. That’s all I used to PFtrack to do his bidding. Later I’d refine track. Put some markers, export to .FBX and voilá: I got my Matchmove.

But I ask this question because when I do the same with my phone tracking is offset. On a Droid Turbo, the resolution of 10MP (4320x2432) can record at 1920x1080 24 fps. The focal lenght is 4.5 mm. Viewing angle 66.5ºx40.3º. Sensor size 5.9x3.3 mm. Crop factor: 5.9x

With those numbers, I don’t know how to setup PF track or any other tracker since:

If I set the sensor size: I get an automatic size frame which I cannot modify (it gets ghosted)

If I set the Image size: I get an automatic crop factor which I cannot modify (to set the sensor size).

I guess this should be pretty straight forward in PF track or any other Tracker (Boujou, Blender, Shiloette, etc…) but I don’t know where to put what information provided above.

Could someone help me with this please?

I’d say forget the numbers, they are only relevant if you want to simulate accurate DOF with lens and camera parameters.

For matchmoving, you most probably don’t know the accurate numbers anyway, whatever values are written on the lens or in manual are not accurate enough to use as hard constraints for solving. Instead, let the software calculate the fov for you based on whatever sensor size. You can leave it a default or set it to something more accurate, makes no difference for camera transformations and solve as long as you don’t constrain both sensor size AND fov. Solver will give you a much better estimate for your fov/focal length than any number written on your lens. If shot is more difficult it is wise to shoot test patterns for accurate measurements of fov (and lens distortion ofcourse, but this should be done for every shot anyway, don’t even think of accurate solve without delensing) that can be used to get a more precise value to be used as solver constraint.

1 Like

I use Blam addon when I can’t get the lens to solve.