The Ship Detection And Recognition Algorithm Based On Polarization SAR Technology Research

Polarimetric Synthetic Aperture Radar(PolSAR) is a new kind of radar which can test thetarget’s polarization characteristics，it has the ability of all-weather、all-day、long-distance andreal-time detection, it can also classify and recognize the target and then it can get more informationof the target’s scattering property scattering characteristics. On the basis of differenttargets’ polarization scattering properties, PolSAR can improve the accuracy of target detection andtarget recognition, so it can test and recognize by extract relevant characteristics of ships.This paper is about the ship’s testing and recognition which mainly based on PolSAR image.On the basis of the information of polarization scattering which is contained in the PolSARimage，this paper’s specific research as follows:(1)Firstly, it’s the pre-processing of the polarization SAR image, its main work isspeckle filtering. This paper introduce the mechanism of speckle’s formation and the model of noise,this paper also introduce3kinds of commonly used filter algorithm: PWF、 Polarization LeeFiltering and Filtering algorithm based on scattering characteristics. These algorithms are testedwith measured data, after comparing the results of tests, it shows that the filtering algorithm basedon the scattering characteristics not only can suppress speckle noise but also can greatly improvethe ability which maintain the image’s detail information. This algorithm’s outstanding feature isprocessing the covariance matrix elements independently, so it has potential advantage than otheralgorithm on maintaining the polarization information, it this way it can provide great convenienceon the further processing of PolSAR image.(2)Because there are already many research of ship testing based on POLSAR IMAGE, thispaper do some primary research on common algorithm of ship testing(K-CFAR, Entropy，OPCGE)，and then put forward a new method of ship detection method--NMF method based on wavelettransform. Firstly, This method do the work of two-dimensional discrete wavelet decompositionof polarization imaging，get the Low-frequency sub-band images and then test the ship by usingNMF decomposition. In this paper, measured data is used in the test, the results show that comparedwith other methods, the new method can be applied to Ship Detection better.In this paper, the proposed method has nothing to do with the ship’s size, there is no need setting thetarget window、background window and protection of the window. At the same time, this methodhas nothing to do with the statistical model of ocean clutter distribution. The computing time is lessthan the traditional CFAR.(3) On the basis of the primary research of existing ship identification method, this papermainly introduce some related kinds of target ship classification method of polarization SAR image, then compare them and analyze these methods’ merit and demerit. In view of thepolarization decomposition method of ship classification recognition experiment, and analyzes theexperimental results.This paper summarizes the existing problem of polarization ship identification.The main problems of Ship recognition is there are some disadvantages in Feature extraction,feature selection, the algorithm stability and assessment of the recognition result and so on。...