|
Automatic processes of manufacturing supervision with digital cameras often need to employ the Scheimpflug condition. This is quite common when registering profiles with a laser light. The Scheimplug condition is applied in this case because usually the registration plane cannot be set in parallel to the plane of a laser profile. Moreover, the low-light conditions forcing the use of large diaphragm size and low image acquisition distance imply that the depth of field is insufficient. Setting up three planes: lens main, image and profile in a way that they intersect within one edge, causes that the axis of the lens impales image plane at a point distant from the principal point, which in the analytical evaluation of the photogrammetric networks is adopted as the best-distortionsymmetry point. The Scheimpflug condition causes that the lens distortion (a feature significantly influencing the central projection) is not symmetrical around the principal point and, assuming this point as the origin of radial rays, leads to significant reduction of accuracy of measurement. A solution to this problem is to include the incidence angle between the detector array and lens main planes in the calibration parameters and their evaluation in the self-calibration network adjustment. This best solution from the substantive point of view thus needs elaboration of a specific software for self-calibration bundle adjustment, which is costly and time consuming. In this paper the different – computationally easier methods for the evaluation of image errors for the images taken considering the use of a camera with the Scheimpflug condition. The first method involves determination of interpolated corrections to measured coordinates based on a “deviation map” obtained from the projective transform of a planar, multi-point test-field on the image. The second method employs the evaluation of the best-distortion-symmetry point using the deviations evaluated after the projective transform, approximated by a radial and tangential distortion polynomial evaluated regarding this point. The research was conducted using real as well as simulated data.
|