Symposium on Remote Sensing for Resources Development end Environmental Management / Enschede / August 1986 
Oil drums as resolution targets for quality control 
of radar survey data 
B.N.Koopmans 
International Institute for Aerospace Survey and Earth Sciences (ITC), Enschede, Netherlands 
ABSTRACT: Spatial resolution in radar imagery is dependent on angular discrimination in along-track (azimuth) 
direction. Range discrimination depends on a time delay measurement, using either the duration of the pulse 
or a frequency-modulated pulse. 
To check effective spatial resolution of Synthetic Aperture Radar (SAR) or Real Aperture Radar (RAR) data, 
measurements can be made on the response distribution of a point target. This method, used for digital data, 
can assess only the peak width to the background and no effect of side—lobes or geometric distortion is taken 
into account. For optically correlated imagery, the method is less applicable. 
A pragmatic approach for control of spatial resolution is to measure the smallest distance between two point 
targets having approximately the same response strength, which are still visible as two objects on the image. 
For this purpose, a number of point targets of materials easily and cheaply available everywhere in the world 
(oil drums cut in half lengthwise) were used in a test site during ESA's SAR-580 experiment in Southern 
Spain. From their separate identification, a good deduction could be made regarding the effective spatial 
resolution in range and azimuth of the imagery. 
INTRODUCTION 
The SARTHI project - (a side-looking radar and IR 
false colour survey over the Iberian Pyrite belt 
(Koopmans et al. 1985) - was part of the European 
SAR-580 experiment carried out during 1981-84. 
During the radar survey flight, an experiment was 
carried out with drum reflectors to check the use 
fulness of a number of point targets to determine 
the spatial resolution of the radar. 
SPATIAL RESOLUTION 
The definition of spatial resolution in its simplest 
form is: "the minimum distance between two point- 
sources (objects) of equal-intensity that a sensor 
can record separately". The definition is not always 
used in this sense, however, and depending on the 
system, other parameters may be taken into consider 
ation. 
Forshaw et al. 1980, recognized four different 
types of identification of spatial resolution: 
1. Geometric properties of the imaging system. 
2. Ability to distinguish between point targets. 
3. Ability to measure periodicity of repetitive 
targets. 
4. Ability to measure spectral properties of small 
finite objects. 
The first type giving the geometric properties of 
the imaging system will be briefly mentioned here 
and compared with results measured according to the 
second type: ability to distinguish point targets. 
The third type, ability to measure repetive targets, 
deals mainly with the optical approach of defining 
spatial resolution in line pairs per milli-meter and 
will not be treated here. 
For side-looking radar, the spatial resolution on 
the basis of geometric properties in range is depen 
dent on the half pulse-length and the depression 
angle; the latter is variable over the image swath: 
Rgr = .. _ 
^ Z cos ts 
where Rgr = resolution in ground range 
c = velocity of light 
t = pulse duration 
3 = depression angle 
Here we are dealing with a range discrimination 
based on measurement of a time delay (duration that 
the pulse is emitted). To improve range resolution, 
the pulse may be frequency-modulated. 
In azimuth, the spatial resolution depends on the 
half-power beam-width and the range towards the 
object. It is an angular discrimination, hence the 
beam-width increases with range. 
The following equation is valid for a real aper 
ture radar: 
where Ra = resolution in azimuth 
X = wavelength 
Rs = slant range to target 
Da = diameter of aperture in wavelengt 
units. 
For a focussed synthetic aperture radar system, 
the azimuth is one-half of the physical length of 
the antenna and is independent of range distance. 
This is theoretical, however, and azimuth resolu 
tion measured according to the second type is often 
larger. 
"Spatial resolution" should not be confused with 
"detectability". Objects smaller than a ground 
resolution cell may be detectable if the small ob 
ject has a strong radiance or backscatter, forming 
sufficient contrast relative to its background. 
Well-known examples of such detectable objects 
are metal fences, corner reflectors, Lunenburger 
lenses, or other radar beacons. 
For the European SAR-580, the effective spatial 
resolution of the SAR imagery was measured on the 
basis of the response distribution of point targets 
(Smith, 1985) (type 4 of Forshaw's spatial resolu 
tion identification). Algorithms were used for 
digital data to select or reject a number of appro 
priate point targets and to calculate the average 
width of the response function at 3 dB below their 
peak value. This method assessed only the peak 
width of the point target to the background and no 
effect of antenna side-lobes or geometric distor 
tion was taken into account. For imagery which is 
not in digital form, the method is less appli 
cable. 
The results of these measurements for data corre 
lated at the Royal Aircraft Establishment (RAE) at 
145