Ship detection of the airborne SAR images

Ship detection is one of the important tasks of seaport surveillance. We develop a series of steps to detect them. First, several filters are used to reduce the speckle. Then the sea image with ships is segmented with the entropic thresholding method. Next, every ship is detected. Finally, we calculate the feature parameters of every ship. Those features can be used for later classification and recognition. In this paper, we compare several methods for each step. In high resolution SAR images, large-size ship targets appear as bright pixel groups with high contrast of the background that depends on the ocean backscatter, varies with incidence angle and sea conditions. The pixels of a certain ship can be divided into two sorts. The high peaks are caused by corner reflection and the in-plane scatterers cause the low peaks. The intensity of the low peaks varies while the imaging conditions alert. So we believe the high peaks would be a more robust characteristic. The airborne SAR is an L-band SAR with a resolution of 1.5 m.