camera calibration

[ahmed KIMOO]

how to adjustment the camera calibration for uav

[macsurveyr]

That pattern is typical of a camera that was set to autofocus. Unfortunately, I have seen it many times.

It also answers your question in another post about Z accuracy. If the camera has been autofocused, the camera calibration will never be as good as it should be. The resulting error manifests it in Z and is typically 10 to 30 times more that it would be if the images were captured at a fixed focus and the camera calibration is fully optimized.

You can still make decent orthos from such projects, but volumes and elevations will unpredictable and should not be depended on.

Tom

[ahmed KIMOO]

That pattern is typical of a camera that was set to autofocus. Unfortunately, I have seen it many times.

It also answers your question in another post about Z accuracy. If the camera has been autofocused, the camera calibration will never be as good as it should be. The resulting error manifests it in Z and is typically 10 to 30 times more that it would be if the images were captured at a fixed focus and the camera calibration is fully optimized.

You can still make decent orthos from such projects, but volumes and elevations will unpredictable and should not be depended on.

Tom

thanks,
there are any method to optimize it?

[spiegel]

That pattern is typical of a camera that was set to autofocus. Unfortunately, I have seen it many times.

It also answers your question in another post about Z accuracy. If the camera has been autofocused, the camera calibration will never be as good as it should be. The resulting error manifests it in Z and is typically 10 to 30 times more that it would be if the images were captured at a fixed focus and the camera calibration is fully optimized.

You can still make decent orthos from such projects, but volumes and elevations will unpredictable and should not be depended on.

Tom

I think, for low-distance-fotos (sculptures, buildings etc) this is very correct.
But I wonder, if it also is important for UAV-Modells.
The ground disatance usually is more then 50 m or so.
Is there any difference from autofocus to fixed focus?

[macsurveyr]

Yes, there is a difference between auto and fixed focus from a UAV. If the camera is set to auto focus the resulting Z accuracy will suffer.

I have investigated many projects and thousands of photos. Whether it is due to slight focus inconsistency or focus breathing or something else, I can't say. I can say the results suffer every time when compared to ground truth or to high quality fixed focus image sets.

Tom

[spiegel]

Hi Tom,

in the landscapes which you compared, have there been any high objects over ground like trees or buildings?
I guess, that in such cases auto-focus could bring a wrong result.

I am not a fotografer, so maybe I am wrong. But I thought, if taking fotos from up to 50 m and more, the focus should be on "infinite" distance. If there are no high objects over ground, why does it make a difference wether the focus is set fixed manualy or set by the camera in automatic mode ?

Frank

[macsurveyr]

Hi Frank,

It does not matter if the landscape has objects or not. Nearly all of the ones I have compared are in fact, very flat areas. It is simply that most cameras will move the lens with every focus attempt. It is clear that they do not return to the exact same position, or at least not all of the lens elements return exactly to the same positions, whether it is very close to the same focus distance or not.

Camera calibration is not trivial, well a very high quality camera calibration is not trivial. Very small differences - tenths of pixels - make a difference. PhotoScan is very capable of high quality camera calibrations - as good or better than any software out there - but high quality image sets at fixed focus are necessary for that to be the case.

I have seen it many times, for many years. Even before PhotoScan existed.

Tom

[spiegel]

Thanks