Geometric Structure Feature Extraction of Ship Target in High-resolution SAR Image

With the improvement of SAR image resolution, ship recognition using SAR image has become a new way to obtain marine intelligence for remote sensing satellite. Geometric structure feature which can be extracted easily is the intuitive description of targets, and becomes an important evidence for preliminary classification of ship targets. This paper firstly discusses some typical geometric structure features, which emphases on the calculation methods and applications for ship recognition. Then a new geometric structure feature extraction method based on azimuth estimation is proposed. Morphological open operation is adopted in pre-processing to eliminate the side lobe, moment technique is exploited for azimuth estimation to rotate the target to horizontal state, and gradual approximation is used for obtaining the minimum bounding rectangle to extract other geometric structure features. The experiments on real SAR images demonstrate that our method can extract the geometric structure features with high accuracy and good stability.