Chemistry Reference
In-Depth Information
for assigning the sub-segments to either the signature of the oil spill or the
residual area. The results of the feature extraction chain are binary repre-
sentations of the associated raw images. These binary images hence are
called
segment maps
.
Fig. 7.
Feature extraction chain (from top to bottom) of the Oil Spill Scene Ana-
lysis System
The extracted spill segments can be easily analysed concerning their
spatial properties. The spatial characterisation implemented in OSSAS is
capable of computing the area, the centre of area, the tilt and the length/
width ratio of each segment. The latter two properties are especially useful
for describing an oil spill in terms of its spreading direction and shape, re-
spectively. They are determined by fitting ellipses to the convex hulls of
the segments. The algorithms used for this purpose are robust and docu-
mented (Barber et al. 1996, Fitzgibbon et al. 1999). Moreover, the fusion
of the TIR and NUV segment maps allows the detection of thermal hot
spots (due to thick oil) and the automated discrimination between thermal
hot spots and ships. This kind of classification is possible because ships are
usually poor reflectors in the NUV. On the basis of modelled and reported
detection limits (Robbe 2005) the extracted thermal and ultraviolet features
are used for the calculation of a minimum volume estimate. This estimate
is part of a decision scheme for volume estimation based on IR/UV-LS,
MWR, IALFS and fused NUV/IALFS information.
