from http://www.ipi.uni-hannover.de/html/mitarbeiter/jacobsen
BACK TO THE LIBRARY of Sagaidak Darija, DonNTU
(fragment)
The georeference of any photogrammetric product is based on the reconstruction of the geometric relations of imaging in a chosen object coordinate system. For the handling of aerial photos traditionally the bundle of rays from the image points over the projection center to the ground points is modelled by means of the camera calibration information and the exterior o rientation determined by means of control points. Today the time consuming ground survey of control points can be reduced by block adjustment or even more by combined block adjustment with projection center coordinates from relative kinematic GPS-positioning. It is also possible to avoid control points like the measurement of image coordinates of tie points by direct sensor orientation with a combination of GPS and an inertial measurement unit (IMU).
Interior Orientation
The interior orientation is required for the reconstruction of the bundle of rays - the location of the projection center has to be known in relation to the image points together with the geometric influence of the lens system. The calibration certificates of the camera manufacturers are including the location of the principal point, the focal length and the radial symmetric lens distortion. This is sufficient for the traditional handling of single models because the remaining differences against the mathematical model of perspective geometry are small. The mathematical model of perspective geometry is expressed by the colinearity equation - image point, projection center and ground point are located on a straight line and the photo is exactly plane. Even if the radial symmetric lens distortion, earth curvature and refraction correction are respected, this is only an approximation.
The tangential distortion, caused by not centric location of the lenses in relation to the optical axis of the lens system is not investigated by the manufacturers and the image is not located in a strict plane surface.
In addition the focal length may be influenced by the air temperature, the air in front of the camera may cause a deformation of the bundle of rays and the photos may be deformed by the developing process. The caused small deviations against the mathematical model may sum up in the case of a block adjustment with only a reduced number of control points, resulting in a deformation of the block. Especially the height is sensitive for such effects.
Photo coordinates cannot be measured directly, with analytical plotters plate coordinates are available and digital plotters are d elivering the pixel addresses of the scanned photos. The photo coordinates are computed by a transformation of the measured to the calibrated positions of the fiducial marks. The film shrinkage, different in both directions and also angular affinity are requiring an affine transformation. A higher degree of transformation is not necessary and is even decreasing the accuracy.
Digital cameras equipped with an area CCD do not have fiducial marks, image coordinates can be computed by a multiplication of the pixel addresses with the pixel size and a shift to the principal point. The multiplication with the pixel size should also respect the definition of the system of axis - pixel addresses dohave usually the origin in the upper left corner and the row is counting downwards, photo coordinates do have the y-coordinates from the principal point counting upwards. Area CCD's do have usually a very stable geometry, not requiring self calibration parameters for the image plane.