ols, Operator- 
(DREO) has 
al-time, high- 
fied as to the 
ime and with 
at a high rate, 
> aids. 
searched and 
(ROICATS). 
face features, 
each of these 
awa a mis au 
de navires de 
itégories. La 
parce qu'elles 
ue graphique. 
érateur radar 
algorithme de 
composantes. 
classification, 
combatant or 
vas introduced 
capability for 
y one of three 
the SPOTIight 
on a specific 
returns arc 
radar image is 
nage, detected 
version of this 
en ocean arc 
motion. In 
  
nonadaptive SPOTlight mode, a high-resolution image 
of a small area of land is generated. This image can be 
analyzed for the presence of man-made targets, and is 
particularly useful for harbour surveillance. 
The second sub-mode is called Range-Doppler 
Profiling (RDP). In this mode the antenna is also 
maintained fixed on a specific target, but instead of a 
single radar image being generated, as in the SPOT 
mode, a continuous series of high resolution radar 
images of ships are generated in real-time. Because of 
the constraints of real-time processing, this mode does 
less adaptive processing than the SPOT mode, but the 
Radar Operator is provided with a greater variety of 
images with time. This mode is not used for land 
imaging. In both RDP and adaptive SPOT modes, 
either stationary or moving ships can be imaged. 
In the third sub-mode, which is called the STRIPMAP 
mode, a continuous strip, or map, of high resolution 
radar imagery is generated. In this mode, spurious 
motion of targets is not compensated, so it is not used 
for imaging ships on the open ocean, where the motion 
of the ship would severely degrade the image. Its chief 
application is for land imaging. 
Because STRIPMAP imagery is generated in real-time, 
a Radar Operator must view a large amount of data at 
a high rate while trying to detect man-made targets in 
the images. If a man-made target is detected, it must 
be analyzed for various attributes, such as its shape and 
location. This analysis can be performed either 
immediately, while still viewing the STRIPMAP 
imagery for the presence of other targets, or more 
likely, by saving an image frame and recalling the 
image for analysis at some future time. Analyzing 
these images can be very demanding on an Operator, 
especially after working for any great length of time. 
To alleviate the potential fatigue that an Operator may 
encounter, some form of aid should be provided to the 
Operator during the analysis process. The aid should 
consist of automated target detection and localization 
procedures, and map files and overlays. 
The remainder of this paper discusses work done in the 
classification of RDP and (adaptive) SPOTlight images 
of ships. 
The appearance of SPOT or RDP images of ships is a 
function of many variables, such as: the location and 
type of scatterers on the target, the SAR system 
frequency, the radar-to-ship viewing angle, the amount 
and type of sea-induced ship motion, and the cross- 
range resolution. The appearance of the images of 
Stationary land targets also varies with the location and 
type of scatterer, frequency, aspect, and cross-range 
resolution. However, they do not experience the large 
changes in the orientation of the image plane which is 
seen with moving ships. Because of the dependence of 
the appearance of a SAR image (i.e. SPOT or RDP) 
on the variables mentioned above, the number of 
possible images that can be generated from any one 
target is large. Since the potential number of targets is 
also large, the total number of possible SAR images is 
very large indeed. It is the job of the Radar Operator 
to understand the abundance of SAR images that are 
generated, and attempt to make a speedy and correct 
decision as to either the type or class of target. Since 
it would be very difficult for an Operator to remember 
all possible images that might be encountered, some 
form of aid should be provided for the classification 
process. 
The Airborne Radar Section at DREO is currently 
developing a computer-based graphics system to aid the 
Radar Operator in performing speedy and accurate 
classification. This system is called the Radar Operator 
Interactive Classification And Training System 
(ROICATS). The present system is composed of three 
main components: a graphical-user-interface which 
supports the use of numerous computer-based tools, 
operator-machine interface features, and automated 
target classification algorithms. The training aspect of 
the system has yet to be developed. 
The tools allow image features to be extracted and 
various forms of information to be manipulated. 
Typical tools include an electronic ruler for measuring 
the distance between image features, an extraction tool 
to extract a subset of an image, and image display and 
manipulation options. Information, such as the physical 
dimensions of a ship, is stored .in an interactive 
database. 
The operator-machine interface provides the structure 
for simplified interaction between the Radar Operator 
and ROICATS. For example, most of the system 
interactions can be done through the use of just a 
mouse or trackball, while avoiding the tedious use of a 
keyboard. 
The automated target classification algorithms analyze 
the SPOT and RDP ship images to determine a list of 
ship classes that may match the target in the image. 
The list should be much smaller than the complete list 
of typical classes, so that the Operator only needs to 
consider a subset of the possible classes. This speeds 
up the classification process and improves the accuracy 
since there is less of a chance of matching the target to 
an incorrect ship class. 
ROICATS is being developed on a SUN SPARCstation 
370GX  engineering-graphics workstation. The 
computer-based tools and operator-machine interface 
325