Marine Ship Target Detection Algorithm For SAR Image

Synthetic aperture radar(SAR)has attracted more and more attention because of its own advantages and imaging characteristics.Nowadays,a lot of research work is carried out around SAR images.Target detection has become the focus of SAR image application.In this thesis,ship target detection algorithms are studied for SAR images with different resolutions.The main work is as follows:(1)Preprocessing of SAR images.In view of the coherent speckle noise of SAR images,the basic principles of several typical filtering methods are introduced.The performance and characteristics of each method are analyzed through comparative experiments,and filtering methods suitable for this article are selected for filtering preprocessing.The filtered image is preprocessed by the maximum between-class variance method for sea and land segmentation,which is used in the subsequent target detection process.(2)Research on ship target detection algorithm for low-and medium-resolution SAR images.The basic principles of CFAR detection are explained,and the basic characteristics of typical clutter statistical models are discussed through fitting experiments.Aiming at the high false alarm rate of traditional CFAR detection algorithms,this thesis proposes a ship target detection algorithm based on Otsu and joint distribution in SAR images.Based on the related joint distribution model,the two-dimensional Otsu algorithm is introduced into the sea clutter statistical model,and the Otsu and joint distribution model are established to simulate sea clutter and improve the accuracy of the statistical model.To solve the problem of too long detection time,this thesis adds a priority decision mechanism to determine whether there is a suspected target in the window,and selectively performs single-pixel sliding or half-window sliding to save detection time.Experimental results show that the proposed detection method has the advantages of high detection rate and less detection time.(3)Research on ship target detection algorithm for high-resolution SAR images.The focus is on the object detection method based on the component model.The compression preprocessing is performed first,and then the points of interest are obtained through the local maximum mark and depth-first search.Aiming at the situation that the error of edge segmentation causes the detection accuracy to decrease,the fuzzy level set algorithm is introduced to extract the target edge contour,obtain the edge pixel and its neighboring pixel information,and establish the connectivity index.Aiming at the distortion of the target shape in the detection results,this thesis uses GLCM to extract texture features in four directions,and combines them with geometric features to reconstruct the similarity index.Finally,based on the geometric constraints between the parts,the existing combinations are screened to generate the description of the ship target and realize the ship target detection.The experimental results show that the improved detection method in this thesis can simultaneously ensure the accuracy of the detection results and the integrity of the target shape.