The Study Of Maritime Ship Targets Monitoring

With the improvement of the synthetic aperture radar(SAR) imaging techniques, SAR images have become more and more widely used in maritime ship targets monitoring. In the application of SAR imagery for ocean surveillance, ship detection and classification is a basic problem to be solved.Statistical modeling is essential to SAR ship detection. It could describe SAR images by statistical methods and reveal the characteristics of these images. First of all, the current researching state of statistical modeling is analyzed, and some classical statistical models are introduced in detail. Finally, a sea clutter statistical method based on similarity fitting is proposed in this paper. We first estimate five classical probability density functions of the sea clutter distribution in synthetic aperture radar (SAR) imagery, which include Rayleigh distribution, lognormal distribution, Weibull distribution, K distribution and G0 distribution. After that, we fit all five models to a new sea clutter distribution by an optimization method based on a similarity criterion.In experiment, four real SAR images are used to evaluate the fitting precision based on the Kullback-Leibler distance. The results show that the proposed method is superior to five classical distribution models. When using the fitting model and CFAR algorithm to conduct ship detection, the proposed method do the best in controlling false alarm and reducing leak detection.At the ship classification stage, a real ship target classification database is used so that the classification results have strong persuasion, and a new ship target classification algorithm based on ensemble learning is proposed. This algorithm first extracts 21 ship features in scale, shape, texture and invariant moments, then output the classification results through ensemble learning classifier. This classifier could combine some weak classifier with weight to form a strong classifier. In ship classification experiment, the real SAR ship samples from the established database are used to classify cargos, containers and tanks. The result shows that the proposed classification algorithm is superior to other two classical classification algorithms.