There are currently five
products that can be generated from the OMW.
-
Vessel
Detection (location, heading and speed),
-
Frontal
Features (ocean current boundaries, eddies, weather fronts)
-
Ocean
Wave Information (length, direction and height)
-
Slicks
(regions of low radar backscatter)
-
Winds
(speed and direction)
Each of these products
produces a data file with information about the variables computed.
Detection of targets (ships) is limited by the spatial
resolution of the sensor, the background clutter levels and the nature of the
speckle noise in the image. Identifying bright point targets and associated
linear features (wakes) is complicated by speckle noise, false targets and
different wake signatures. Wakes manifest themselves in several forms in SAR
images including Kelvin waves (bright), dark turbulent wakes, stern wakes and
internal waves generated by the ship's passage.
The OMW employs a two step process for ship and wake
detection. The system first carries out a target detection procedure to identify
candidate vessels, followed by a procedure for detecting and analyzing wake
signatures.
The target analysis employs two algorithms. The first, N
sigma, divides the image into smaller areas to minimize large scale backscatter
variations. The local mean and standard deviation is calculated to determine a
target threshold above which a target is considered real. The second algorithm
is a Constant False Alarm Rate approach using a K-distribution to model the
image intensity probability function.
The wake analysis is carried out in the Radon transform
domain where linear features in the original image appear as peaks or troughs. A
method known as the Dempster-Schafer algorithm makes use of belief functions to
assist in determining if a peak or trough is related to a wake feature. In cases
where the wake is identifiable, the vessel speed and heading is also computed
based on the offset of the ship location from the apex of the wake in the along
track satellite direction.
This product shows features related to oceanic and
atmospheric fronts as well as eddies that are extracted from the SAR images. The
processing uses image morphology to isolate significant features and these are
then classified by a comparison to averaged and normalized profiles of pixel
values across the detected feature to ideal profiles for each type of feature.
Results are expressed in terms of strings of latitude and longitude points that
constitute individual features.
This is a spectrum from a sub-image extracted from the
larger SAR image. The spectrum is a measure of the image pixel value variance.
As such it is not an ocean wave spectrum, and is not scaled to represent a power
spectral density for the ocean surface wave field. In order to do so, a complete
model for the radar backscatter from the ocean surface must exist., which
describes the modulation in backscatter associated with the ocean surface wave
field. This model must be invertible to be solved for the ocean surface wave
spectrum. Due to the non-linear imaging mechanisms that contribute to the wave
patterns observed in SAR images, such a model does not exist.
Coupling the estimate of the wind, with the estimate of the azimuth
cut-off
wave number permits a determination of the significant wave height to within
approximately 0.5m.
There are no graphical products generated for the wave
information derived from the SAR image.
This extracts regions of exceptionally low backscatter from the image which
can be caused by biological films or oil spills. The output consists of
information both as a map and a table of the location, shape and spatial extent of
such features.
The wind product relies on calibrated SAR data to extract radar cross section
values and estimate wind speed. Under suitable circumstances, the wind direction
may also be estimated use low frequency variations in backscatter over the
ocean. The output consists of a graphical display and a table of wind speeds and
directions on a grid superimposed on the image.