Research On Sea Target Detection Method Based On SAR Image

Synthetic Aperture Radar(SAR)has a wide range of applications and good imaging performance.However,in sea-level scenes,due to the complex and varied sea clutter,it is very difficult to analyze the characteristics of sea clutter in SAR images and perform target detection on it.At the same time,due to the development of marine applications,countries have paid more and more attention to protecting their own maritime rights and interests in recent years.Therefore,how to detect ship targets in SAR sea images has become a research hotspot.Based on the actual situation and needs,the thesis studies the related methods in three aspects:SAR image preprocessing,maritime target detection and maritime target recognition.These three aspects are used as three important steps to form a complete method of ship detection and recognition processing.In the aspect of SAR image preprocessing,a wavelet denoising method based on stationary wavelet transform and a sea-land segmentation method based on the Otsu method are proposed.Because it is hardly to get guided image in the traditional guided filtering denoising method and the traditional wavelet denoising method may lost detailed information in the wavelet transform process,this paper improves the wavelet denoising method and uses the stationary wavelet transform in the transform process.So that the image information is kept as much as possible and no guided image is needed.After denoising,this paper based on the Otsu method to carry out sea and land segmentation to remove terrestrial interference.In the aspect of maritime target detection,a cascaded CFAR detector based on improved background window is proposed.In the thesis,the characteristics of sea clutter are analyzed and modeled to analyze which model is suitable for SAR images under sea conditions.Then CFAR detection is performed based on sea clutter model.For the traditional CFAR detector,it is difficult to achieve uniformity in accuracy and efficiency,and the background window is easy to be interfered.This paper first performs global CFAR detection on the image,and then performs local CFAR detection through the improved two-layer background window.The model estimation of the sea clutter background is more accurate,and the detection eff-iciency is effectively improved without reducing the detection accuracy.In the aspect of maritime target recognition,an improved SVM target recognition classifier based on unequal distance optimal hyperplane is proposed.Because of the constant false alarm theory,the detection results often have false alarm targets.Therefore,based on the effective data set,the paper firstly preprocesses the slice target,extracts the geometric features of the ship target,and then designs the target recognition classifier for classification and recognition.Due to the poor effect of the traditional minimum distance classifier for identifying false alarm targets and the disadvantage of the traditional SVM classifier that does not consider the training sample distribution,this paper improves the optimal hyperplane position of the SVM classifier so that it is not equidistant from the positive and negative target intervals,which improves recognition accuracy.After the above three aspects,it can effectively process,detect and identify the ship targets in SAR image,comparing with the traditional method.Finally,the method used in this paper uses SAR satellite image to simulate in Matlab environment.