Ask Your Question

Revision history [back]

click to hide/show revision 1
initial version

stereoCalibrate instability with 3D artifact

I get poor and unstable results performing a stereo calibration against a 3D object. Filtering outliers makes the error alternate between mediocre and extremely high.

If I resubmit the data only with points on the object's bottom plane, and do the same type of filtering, then the stereo solution is well-behaved and satisfactory.

The cameras are not parallel, but I calibrate their intrinsics separately. I'm using 2.4.3. Please advise what is happening here.

Source files

Planar Trials
Error: 0.747405
Error: 0.745686
Error: 0.73781
Error: 0.758419

3D Trials
Error: 1069.91
Error: 29310.4
Error: 5.67888
Error: 461.579
Error: 5.39043
Error: 603.938
Error: 6.34686
Error: 5.95827
Error: 2406.1
Error: 7.11539

End