296
S. Lehner et al.
Fig. 15.11 Ship detection
work flow scheme
generally difficult due to inaccuracy of recorded coastline, tidal variations, and
coastal constructions.
• Prescreening: applying a simple moving window adaptive threshold algorithm to
detect bright points. In general classical CFAR algorithms are chosen due to its
robustness. Details of other ship detection methods can be found in (Crisp 2004).
• Discrimination: rejecting some false alarms using target measurements or
characterization of specific oceanographic or meteorological phenomena.
• Feature Extraction: Based on standard SAR intensity images ship length, width,
and heading is automatically extracted. If the Displacement between ship wake
and the ship is detectable, this feature can be used as a direct measurement for
ship speed retrieval.
Several algorithms have been developed to detect targets in SAR imagery. CFAR
algorithms have been applied widely among them. However, CFAR and its improved
algorithms have to build a statistical model for the clutter. Since there are a number
of factors influencing the clutter, for example, the type of the ground, it is difficult
to build an accurate model to simulate the clutter.
The CFAR is an adaptive threshold algorithm. Its aim is to search for pixel values
which are unusually bright compared to those in the surrounding area. This is done
by setting a threshold which depends on the statistics of the area under analysis. The
pixels which lie above the threshold are selected as target samples. If the threshold
is chosen so that the percentage of background pixel values which lie above the
threshold is constant, thus the method is called constant false alarm rate detector.
S. Lehner et al.
Fig. 15.11 Ship detection
work flow scheme
generally difficult due to inaccuracy of recorded coastline, tidal variations, and
coastal constructions.
• Prescreening: applying a simple moving window adaptive threshold algorithm to
detect bright points. In general classical CFAR algorithms are chosen due to its
robustness. Details of other ship detection methods can be found in (Crisp 2004).
• Discrimination: rejecting some false alarms using target measurements or
characterization of specific oceanographic or meteorological phenomena.
• Feature Extraction: Based on standard SAR intensity images ship length, width,
and heading is automatically extracted. If the Displacement between ship wake
and the ship is detectable, this feature can be used as a direct measurement for
ship speed retrieval.
Several algorithms have been developed to detect targets in SAR imagery. CFAR
algorithms have been applied widely among them. However, CFAR and its improved
algorithms have to build a statistical model for the clutter. Since there are a number
of factors influencing the clutter, for example, the type of the ground, it is difficult
to build an accurate model to simulate the clutter.
The CFAR is an adaptive threshold algorithm. Its aim is to search for pixel values
which are unusually bright compared to those in the surrounding area. This is done
by setting a threshold which depends on the statistics of the area under analysis. The
pixels which lie above the threshold are selected as target samples. If the threshold
is chosen so that the percentage of background pixel values which lie above the
threshold is constant, thus the method is called constant false alarm rate detector.