Once the distortion, "d", is known, the x and y components are 
computed by: 
dx = (rx/dist)*d and dy = (ry/dist)*d 
where "rx" and "ry" are the x and y components of the radial distance and 
"dist" is the radial distance. 
Atmospheric refraction is also handled via an interpolation 
table. Here the entries are generally handled as a scale factor required 
to correct the vehicle-to-ground direction cosines. The scale factor is 
expressed as a function of the cosine of the nadir angle. 
The calibrated focal length values must be supplied. 
THE TIME-DEPENDENT PARAMETERS 
The following parameters are considered, in a general sense, to 
be time-dependent: 
1) Instantaneous vehicle position (generally in a Local 
Tangent Coordinate System (LTCS)). 
2) Instantaneous attitude components of roll, pitch, and yaw 
at the optical center (note that this is essentially the 
combination of vehicle attitude and scan angle for 
panoramic systems). 
3) Principal point offsets. 
4) Sine and cosine of slit- or film-skew angles. 
The time-dependent parameters are collected into a "vehicle 
table" for each frame of interest with a "row" in each table consisting of 
the values of the parameters for a corresponding instant in time. Figure 3 
illustrates this concept. 
FILM COORDINATES AS A BASIS FOR TIME 
Time in the numerical model is represented by the y-film 
coordinate. The module which generates the vehicle tables must compute a 
bracketing set of y-film values which encompasses the area of interest on 
the imagery. This encompassing set is then subdivided into an appropriate 
number of equal, discrete intervals at which the parameters are evaluated 
using the actual sensor geometric model. 
For the inverse problem (computing image coordinates from 
ground coordinates), one is faced with a standard problem of dynamic 
imagery. In order to know the taking geometry of the camera system, one 
must know tíme, but time is only known after the image position is known, 
which depends upon the taking geometry of the photograph. This type of 
problem is typically solved by an iterative solution, and the same is true 
in this case. We have an advantage, however, in that we have easy access 
to some frame approximation. We could pick any time within the frame, 
578