Discrimination of Ship’s Bow in SAR Image

Abstract

Discrimination of ship’s bow is a great application for marine surveillance in synthetic aperture radar (SAR) image. The real ship contour is approximately elliptical and the bow’s curvature is significantly larger than the stern’s curvature. The increasing resolution of SAR image makes ship target appear to be more structured and shaped which makes discrimination of ship’s bow possibly in SAR image. In this paper, we advance a method for discrimination of ship’s bow which includes two important parts. The first part, we use the gradient vector flow (GVF) snake model with elliptical constraint (EC) to get a more realistic ship contour for curve fitting. The second part, we use the least squares quadratic polynomial curve fitting to fit the two ends of the contour along the ship’s main shaft. The bigger absolute value of second order polynomial fitting coefficients, the more curved the rigid curve. The more curved end is identified as the ship’s bow. Experiments performed on detected small-size ship slices which also called regions of interest (ROI).These ROI slices which come from RADARSAT-2 data and one ROI slice contains only one ship target. Experiment results illustrate the effective performance of proposed method on discrimination for the ship’s bow in SAR image.