Two basic approaches for a geometric correction on a pixel-by-pixel basis are distinguished: (a) A parametric 
approach using the location of the airplane and inertial navigation system data to simulate the observation geome 
try and (b) a non-parametric approach using tie points or ground control points (Itten and Meyer, 1993). It is well 
known that the non-parametric approach is not reliable enough for the unstable flight conditions of airborne sys 
tems, and is not satisfying in areas with significant topography, e.g. mountains and hills. The present work 
describes a parametric preprocessing procedure which corrects effects of flight line and attitude variation as well 
as topographic influences and is described in more detail by Meyer (1994). 
2.2. Basis of the Study 
Test site and image data: The area “Zug-Buochserhom” is the standard test site of the Remote Sensing Laborato 
ries, University of Zurich-Irchel in Central Switzerland. The region was covered by the AVIRIS flight #910705, 
run 6 of the NASA MAC Europe'91 campaign providing a data swath with an average nominal pixel size of about 
18m. The first scene, Zug, represents a hilly area with highest elevation differences of about 600m and slopes with 
typical angles between 15° and 60°. The second scene, Rigi, is an example of mountainous terrain with elevation 
differences of about 1400m and maximum slope angles up to 90°. Figure 2 shows a to-scale-profile through the 
mountain Rigi to give an impression of the topography and especially of the slope angles. 
Horizontal distance [m] 
Figure - 2: To-scale-profile through the mountain Rigi to describe the topography. 
AVIRIS auxiliary data: The quality assessment for the actual data set is described in detail in Meyer et al. (1993a). 
The current work uses the navigation data roll, pitch, and true heading (generated through the ER-2 Inertial Navi 
gation System) and the roll and pitch of the AVIRIS instrument’s precision gyros. 
Digital elevation model (DEM): The test area is covered by the two digital models (DHM-25) Zug and Rigi gen 
erated by the Swiss Federal Office of Topography 1 . They have a resolution of 25m in i and j direction and of 0.10 
m in elevation e with an average error in elevation of 2.2m±1.0m for model Zug, and 4.4m±1.8m for Rigi. 
ADOUR conical radar tracking system: The ground-based immobile tracking radar system ADOUR is a dual 
antenna, dual frequency radar with a conical scan tracking system operated by the Swiss Air Force (Thomson- 
CSF, 1987). For the current approach the three parameters latitude x, longitude y, and altitude z are used. The sys 
tematic error for elevation and azimuth is ± 0.2mrad and ±7m for distance with an update interval of 0.2 second. 
Ground reference information: A forest map was generated by scanning the green (forest) plate of the Swiss 
Topographic Map, scale 1:25,000, edition 1987 at 50 |im with a Optronics 5040 Scanner. The average carto 
graphic accuracy is about 5.0m. A shoreline map was produced by digitizing the same map in an ARC/Info using 
a digitizing tablet. The average (theoretical) accuracy is ± 8.0m. 
2.3. Method 
The method is described in more detail by Meyer (1994). The basic goal is to reconstruct for every pixel the geo 
metric situation at the time it was acquired with AVIRIS. This includes three major aspects. The first considers the 
flight line and attitude of the ER-2 aircraft, the second reconstructs the current observation geometry and the third 
treats the situation on the surface. This approach includes the three different coordinate systems (c,r) for the raw 
file, longitude x, latitude y, and altitude z together with roll со, pitch ф, and true heading % for the observation 
geometry and (i,j^ f p for the DEM. 
l.DEM data courtesy: Swiss Federal Office of Topography, January 24, 1994