113 
parametric 
pn geome- 
. It is well 
borne sys- 
sent work 
on as well 
Laborato- 
#910705, 
:e of about 
lopes with 
i elevation 
irough the 
J. (1993a). 
:rtial Navi- 
1 Rigi gen- 
tnd of 0.10 
igi. 
1 is a dual 
Thomson- 
d. The sys- 
2 second, 
the Swiss 
rage carto- 
flnfo using 
el the geo- 
msiders the 
id the third 
for the raw 
pbservation 
Flight line and attitude of the ER-2 aircraft: The x,y, and z of the aircraft needs to be known as a first step. For the 
1991 European flight the information results from an unaided LTN90-116 INS navigation system with a position 
accuracy of 0.9 nmi/h (Perrin, 1993). These data are not accurate enough for the current approach. Therefore, 
ADOUR radar mesarurements are used as an alternative. The description of the attitude of the aircraft is based on 
the x from the navigation data and the to and $ from the instrument data. 
Current observation geometry: The basic idea is shown in Figure 3 and described in more detail by Larson et al. 
(1994). 
Aircraft leveled Aircraft rotated 
Figure - 3: Observation geometry: The left part shows an ideal situation with plane being leveled (roll to = 0 , 
pitch x = 0°, and yaw e = 0 °. The right side demonstrated the influence of attitude changes related 
to co, x, 0 * 0 . 
The effort is to obtain the underlying surface out of the well-known location (=flight line) and the current attitude 
of the aircraft. The position vector x c Represents the location of pixel (c,r) in the aircraft coordinate system (x,y,z) 
at the instant the pixel was acquired by the instrument and for the ideal case where p = <(> = x = °° (Figure 3, left 
side). 
[( maxP - FOV~\. 
[,an Ll c —2— 
( 1 ) 
where c= pixel number of pixel (c,r) within line r of the raw image, maxP =maximum number of pixels per line 
(614), FOV=Field of View (in rad), and z x =altitude of ER-2 for the current x and y. The pixel wise calculation 
of the actual pointing direction includes correction of the panoramic distortion. The vector X c r is modified by 
rotations about the to, <j>, and x axis (vector X cr Figure 3, right side). The transformed position vector X cr is com 
puted as follows: 
x" cr 
l -x -<t> 
X 1 -to X c r . 
<(> to 1 , 
( 2 ) 
Situation on surface: The topography causes a shift in the apparent pixel location, and affects the pixel size. The 
goal is now to find the intersection between the pixel location vector x i; . and the surface of the DEM. Within the 
neighborhood around the transformed pixel location vector x j; ., a test vector X’j is searched for, which converges 
to X,