Camera solve for dji magic drones

To get an accurate camera solve we obviously need lense information from our camera.

When this information is limited. Like with the mavic mini.
Lens info I can find gives sensor size, but not the actual lens size. Just labels it as 24mm equivalent.

Does this mean we can put a 35mm sensor as sensor size and 25mm focal length and get an accurate solve?

Or do we need to put the crop sensor size in and work out the true focal length?

Cheers for any information

To get an accurate camera solve we obviously need lense information from our camera.

This is only partly true. Depending on what you're planning to do, a completely automatic solve will give you an accurate enough result. The only thing that really matters for matchmoving is that the angle of your solved camera matches the angle of the original camera. (this is how forced perspective shots are created with models etc - as long as the angles match you only have to adjust for size)

So what is your goal?

Thanks for the info.

At the moment I am just doing freelance work, putting camera tracked text into scenes. So as long as it is close its OK.

But I am also self learning pftrack, and want to move onto super accurate matchmove and object tracking. Then eventually learn 3d model composting.

If you have any resources you can point me to, that explains how to calculate the correct lens and fov details I would be most appreciative.

Thanks

The most accurate lens data is obtained by solving first the lens distortion using test pattern or better yet, test sequence (distortion solving removes the curvature and allows filling the pinhole camera model assumption of matchmove solver) and then finding the lens fov by solving a scene with precisely known properties. It can either be a scene which contains calibration objects (Syntheyes has some funny tutorials with balls with sicks) or some exact survey data obtained by precise photogrammetry or scanning. Reason for exact data is to constrain the solver in a way that it must produce accurate fov value, not take a shortcut.

One simple solution is to take a scene with good layout of well defined objects and shoot it with a strong parallax difference in shot, getting at least 90 degree angle between extreme positions. This also forces solver to get accurate fov because any depth miscalculation will be constrained by cross views.

I think i understand the basis of what you are saying.

Like I said I am still new to and learning camera tracking.

I will keep educating myself on understanding and calculating distortion

You might be interested in exactly this video since you're talking drones and tracking in fusion

You might be interested in exactly this video since you're talking drones and tracking in fusion

Video is not working

Maybe it will work like this?