Oil slicks detection processor
Oil slicks detection processor is intended for oil slicks detection against a background of homogeneous sea surface.
Input data are the radar images from spaceborne synthetic aperture presented in some extern data format (CEOS,
XML and so on).
The result of the processing is raster binary image with detected slicks (mask) and set of the slicks parameters
(square, geographic position). Result of the processing may be exported to some extern data format (for example,
to graphic format Geo Tiff), which allows to save georeference data for output image.
Approaches and processing technologies which are realized in Oil slicks detection processor allows user to get
result automatically. But in the case of complicated scenes user has possibility to take part in processing via
adjustment of some processing parameters. Quality of the result may be increased by means of parameters variation.
Processing flow embodied into Processor concludes a few steps: selection of interested area from whole input
scene, semiautomatic image searching and segmentation, detection of slicks via classification procedure, filtering
of retrieved objects, calculating of output statistics on defined slicks. The flow chain settings of processor
could be saved as the project form for possible subsequent application.
||The initial data for the processor is a SAR image, presented in the
some extern format (CEOS, XML).
The user interface is designed according to the processing flow:
Subset of interested region;
Basic operations of the Oil slick detection Processor work flow diagram
An aim of adaptive thresholding is to get an initial classification. This initial
classification is realizing by thresholding. The threshold value is calculated locally for
each pixel of the input image and it depends on the statistical characteristics of the image
inside some neibourhood of the pixel.
Iterative classification works in the feature space and with the input image simultaneously.
This approach allows to get spatially coherent result of classification. Classification problem
is vital for the optimization energy function problem.
An aim of the iterative filtration stage is to delete small pixel’s groups,
which are gather rounded by contrary class pixels.