Show Posts
1
Camera Calibration / Evaluating lens calibration quality
« on: October 03, 2022, 09:34:56 PM »
I'm trying to calibrate a camera lens using MetaShape's chessboard calibration system, and I have a few questions:
1) Is there any numeric indication of how good/uncertain a derived camera model is? I've found a very old post that refers to getting "lower Std Error" as they increased the number of images in their calibration dataset, which I suspect was an output in the old Lens software. I can't find anything similar in MetaShapePro 1.8.3.
2) When using the Tools->Lens->Calibrate Lens... tool, Metashape displays red and blue lines on each image. Are these the residual errors for the corner positions, that is, if the corner identification and lens model were 100% perfect, would they all be expected to be 0?
3) If the answer to the first two questions is "no", is there any way to evaluate whether the lens calibration process produced a viable result?
1) Is there any numeric indication of how good/uncertain a derived camera model is? I've found a very old post that refers to getting "lower Std Error" as they increased the number of images in their calibration dataset, which I suspect was an output in the old Lens software. I can't find anything similar in MetaShapePro 1.8.3.
2) When using the Tools->Lens->Calibrate Lens... tool, Metashape displays red and blue lines on each image. Are these the residual errors for the corner positions, that is, if the corner identification and lens model were 100% perfect, would they all be expected to be 0?
3) If the answer to the first two questions is "no", is there any way to evaluate whether the lens calibration process produced a viable result?
