Camera Calibration

Thank you all for the responses. Im not sure what Pix4D has to do with this but maybe you can explain that further.

Oto, the photos are undistorted and the grid lines are very close. This camera is part of an aerial Lidar system that we process using Teledyne Optech's LMS software. LMS converts the images from the PhaseOne format (*.IIQ) to tif so we can process in Context Capture. Im not sure if the original images are distorted before LMS converts them. We don't currently have any software to view the original format. I have attached the calibration report so you can see the difference in values from the calibration that came with the camera from PhaseOne to the values that Context Capture finds. Why are they so different?

Sylvain, I understand how to hold the calibration, which does not produce good results (Reprojection RMS of 1.48 pixels compared to 0.59 when adjusted), but when I use adjust, it seems to ignore any values I have entered and only use those values as a comparison in the report. I get very close to the same results from Context Capture no matter what values I enter to start with.

As the photos are already undistorted then the calibration is not really needed. Values which you entered from PhaseOne are for distorted original photos so are not kept after aerotriangulation. But parameters are not that much different notation e-5 means that the number is scientific notation as it is too small or too large to display so it is shown as exponent. 1.49e-05 is real number 0.0000149 so almost zero.

Try to run aerotriangulation with calibration parameters K an P set o 0 and option "keep calibration". 

Thank you again Oto. I understand what the "e" stands for and even though I know very little about camera calibrations, just looking at the numbers, they seem to change quite a bit to me, relatively speaking I guess. But I think the bigger issue is the distortion. I didn't really think of that until I typed the previous response. And when I put all zeros in for K and P and hold it, it comes out almost the same as when I hold the actual calibration values. 

Thank you again Oto. I understand what the "e" stands for and even though I know very little about camera calibrations, just looking at the numbers, they seem to change quite a bit to me, relatively speaking I guess. But I think the bigger issue is the distortion. I didn't really think of that until I typed the previous response. And when I put all zeros in for K and P and hold it, it comes out almost the same as when I hold the actual calibration values. 

Surprise isn't it? The calibration seems unneded for already undistorted photos and Contextcapture even gets worse results doing calibration for already unwarped photos. This is where it seems another SfM sotware Pix4D has made it more proof for starters. 

So I found out the images are still distorted when they come out of LMS, there is just very little distortion on these lenses. That means the calibration should work. Do I hold the calibration or let Context Capture adjust it to something different? Am I entering in the correct values? Like I said originally, there are 3 calibrations that came with the camera and the Australis seems like the correct one but I dont know for sure which one to use. Seems like a good calibration should make the results better not worse. 

Try with zeroing and also try with letting CC do the calibration.

What is your acquisition plan? If you do not make a 3D calibration around an object then the calibration may not be accurate.

I have done both and letting CC do the calibration always gives me better results. The calibration I enter is from the manufacturer and came with the camera. You probably know this but these are really expensive medium format cameras. Is that the wrong calibration to use? Is there another calibration we should perform? 

Thanks Oto!

Yes you should precalibrate camera using automatic calibration and only then compare it to manufacturer - https://communities.bentley.com/products/3d_imaging_and_point_cloud_software/w/wiki/28536/the-model-appears-to-be-curved-instead-of-being-horizontal-after-at

Each software has slight differences how this calibration is implemented so the manufacturer parameters could be used mainly as initial values and let software adjust them. 

Which phase one camera you use?