OW 
O0 ® oO eto rro 
D + 0) DM 
provides the altitude element. The inverse problem refers to the 
computation of film coordinates as a function of ground point position. 
When dealing with dynamic sensor models, the inverse problem 
becomes a computationally intensive and time-consuming operation; 
unfortunately, this problem is encountered more and more often in current 
photogrammetric applications because of the need to compute film and pixel 
positions to support graphics superimposition. Because one has no 
knowledge of the time of imaging (i.e., film coordinates) for a given 
ground point and a given frame, an initial estimate must be used to 
generate the time-dependent parameters and then an iterative scheme 
employed to obtain the correct answer. 
If numerous points are being computed, as when digitizing 
points along a road, the same framelet may be used for several consecutive 
point computations. Multiple use of the same framelet lessens computation 
time without degrading accuracy. Alternatively, the framelet computation 
could be computed in a background task, and the information updated on a 
timed interval, so that current framelet information at the point being 
processed is constantly available. 
PARAMETERS OF THE NUMERICAL MODEL 
The parameters of the geometric model can be broken down into 
the elements of interior and exterior orientation. However, for the 
numerical model it is more convenient to discuss the parameters in terms of 
static and time-dependent parameters. 
THE STATIC PARAMETERS 
We typically define the relationship between a mensuration 
device (such as a comparator or analytical steroplotter) and the sensor- 
dependent film coordinate system to be an affine  (six-parameter) 
transformation. The film coordinate system must satisfy the following 
rules: 
1) The y-axis is oriented in the direction of "time". For 
example, in a panoramic model this is in the direction of 
the scan motion. 
2) The y-axis coincides with the optical center of the sensor 
(it is on the "center trace"). 
3) The x-axis is orthogonal to the y-axis. 
We further define image coordinates as film coordinates 
corrected for principal point offsets and compensated for lens distortion. 
Such coordinates, however, are internal to the various comput ational 
algorithms of the numerical model itself. 
Radial lens distortion is modelled by an interpolation table 
consisting of distortion values as a function of the radial distance from 
the principal point. The values are expressed at discrete intervals 
specified by the user. 
577