the geometric accuracy of four Seasat SAR images covering the same piece 
of ground along the banks of the River Tay in Scotland including the two 
cities of Dundee and Perth. Each of those images had been processed 
differently. The tilted-plane optical processor owned by the Enviromental 
Research Institute of Michigan (ERIM) was used in two different models 
to process the data for this test area. In the survey-processing mode, 
the optical correlator elements of the processor are set to certain mean 
values of the satellite attitude and position on the basis that they 
are fixed for the whole path. In the alternative approach of precision 
processing, the optical correlator settings are maintained for only certain 
limited portions of the SAR swath in the along-track direction, the purpose 
being, it was hoped, to improve the spatial resolution of the image. 
  
The Seasat SAR data processing by RAE at Farnborough was carried out in 
two stages. Stage one consists mainly of transferring the original SAR 
data from the original high density digital tapes (HDDT) to a 300 Mbyte 
disc using a Prime Computer; and again from the 300 Mbyte disc onto Computer 
Compatible Tapes (CCTS) using the same Prime Computer (e.g. see Lodge, 1980). 
In stage two, a Prime h00 equipped with an FPS AP-120B array processor 
is used. After checking and correcting the input data from stage one for 
missing data and random errors, the range compression is carried out by 
Fourier Transform techniques using the array processor. Next, azimuth 
compression (beam-sharpening) and radiometric and geometric corrections 
were carried out, after which image pixels were sorted out and an image 
tape was produced in a line-by-line format. 
The digital processing of Seasat SAR data at DEVLR (Oberpfaffenhofen, FRG) 
was carried out in a manner similar to that at RAE, Farnborough. A fast 
32-bit Interdata 8/32 mini-computer equipped with an FSP array processor 
was.used.The software was originally developed by Macdonald-Detweiler 
Associates (MDA) in Canada and was then purchased by ESA as a proven off- 
the-shelf item which would allow immediate implementation of Seasat SAR 
digital processing. 
DESCRIPTION OF THE TEST AREA AND MATERIAL 
The test area used in the present experiment for the metric accuracy 
testing of Seasat SAR imagery. has topographic relief ranging from sea- 
level to around 350 m. The terrain is highly varied with some urban areas, 
a great deal of flat and undulating ground,and some quite hilly areas. 
The root-mean-square relief variation as computed from identified ground 
control points on the images is of the order X 75 m. 
Points identified on the SAR images and the topographic maps of the area were 
selected and their coordinates were scaled off the 1:63 360 Ordance Survey 
maps. However, it must be emphasized that positive identification of those 
points was very difficult on all images; though to a lesser extent on the 
two digitally processed ones. As an example, road intersections which 
might be considered as most suitable for this purpose, proved singularly 
difficult to identify with certainty. A result of this is that only 
definite and unambiguous examples were used. Water features posed their 
own problems in this respect; often being ill-defined and discontinous on 
- 29 -