3.2.
84
Evaluation of Geometric Quality
More recently the problem of the assessment of geometric
quality has been considered, The matching of a SAR image to a
map presents difficulties because the radar locates a target in a
fundamentally different way from a cartographer or optical
sensor. In a SAR, systems imperfections can cause subtle,
unpredictable and time-varying changes in the image geometry.
The aim of this phase was to develop a technique of
characterising distortion which embodied the physics of the
general SAR system, and provides measures which give insight into
the causes of any distortion.
Geometric distortion has three basic causes
i) terrain variation,
ii) uncorrected platform motion and system instability, and
iii) processing algorithms
Much work has been carried out on correcting for terrain effects,
and systematic distortion, apparently little on time-dependent
distortions.
As the main guiding principle, the range-azimuth image
coordinate system provides a natural frame in which to work, and
this should be reproduced in the measurment frame to allow
range and azimuth effects to be treated separately. In addition,
measures of distortion should be insensitive to small operator
errors in target location.
The following indicates the main steps in the procedure:-
À + Locate a set of targets, spaced in azimuth, in bands at
near and far range of the image, and locate them on a map;
2. Use the near range points to define a mean flight
direction;
3. Transform map into coordinate system defined by mean flight
direction;
4, Fit the image to map points by local scaling and
translation only, at both near and far range in transform
image.
Four scale factors are set up which relate to the
dimensions of an image pixel at near and far range. These scale
factors are used to calculate mean terrain slope, azimuth
dependent skew of image lines, and an overall skew, for example.
Software generated to carry out this scheme has been
incorporated into the image analysis package at ESRIN.
