Ship Target Detection And Discrimination Based On SAR Image

Synthetic aperture radar(SAR)is a high resolution imaging radar with the capability of working in all-weather and day-and-night conditions,which is becoming the focus of attention of the researchers in many fields.Considering the national status with massive marine resources,it is of great importance to pay more attention to the ship surveillance in the marine area of our country.And SAR is one of the important techniques for maritime remote sensing.With the development of radar imaging,lots of SAR images are acquired,which provides a good foundation for ship detection and discrimination in SAR images.This thesis focuses on the ship target detection and ship target discrimination,composing of sea-land segmentation,ship detection,and ship target discrimination.The main work of this thesis is summarized as follows:1.Two kinds of sea-land segmentation techniques are summarized.One is based on the global threshold,and the other is based on the superpixel-based multi-level local pattern histogram(MLPH)method.The three groups of measured data sets are used to certify the effectiveness of these two kinds of sea-land segmentation techniques.In addition,the merits and demerits of these two techniques are also analyzed.2.The ship detection and discrimination techniques are introduced.Firstly,we introduce the two-parameter constant false alarm rate(CFAR)detector and the maximally stable extremal regions(MSER)based ship detector.And the measured data sets are used to verify the performance of these two methods by changing the signal-to-noise ratio and the signal-to-clutter ratio.It should be pointed out that the clutter of high intensity is detected because the above two methods heavily rely on the intensity information.To address the aforementioned problem,the SAR imagery adapted locally adaptive regression kernel(SAR-LARK)based ship detector is studied.In the first step,the SAR-LARK features corresponding to each pixel in SAR images are extracted;then,based on the features extracted,a non-parametric kernel density estimation method is performed to compute the saliency map on which the local maximum detection is imposed;in the final step,the maximums greater than a preset threshold are assigned to ship targets,and others the clutter.This method jointly utilizes intensity and local structure information of the image,which makes the detection rate Pd be higher and the false alarm rate Pf lower.Finally,this chapter simply introduces the traditional SAR target discrimination features and the way of feature separability analysis.And the performance of these features in ship discrimination is verified by experiments.3.Software on SAR ship detection and discrimination is designed.Matlab programming technology is used to embed SAR ship detection and discrimination algorithms into a human-machine interface,which makes the process visible.And the whole software includes preprocessing,detection,and discrimination modules,and can complete the function of ship detection and discrimination.