Research On Ship Target Detection Algorithm In SAR Image Based On Information Geometry

The detection of naval vessels on the sea surface holds significant importance for maintaining China’s maritime rights and ocean governance.Synthetic Aperture Radar(SAR)has many advantages,such as operating around the clock in all weather conditions and providing continuous coverage,making it a high-resolution microwave imaging radar that can work uninterrupted.Therefore,detecting naval vessel targets from a large number of high-resolution images obtained from satellite-borne and airborne SAR has extremely important research significance.The main research content of this thesis is based on information geometry methods,which proposes an improved algorithm corresponding to the shortcomings of traditional target detection algorithms.This thesis first describes the development history and research significance of Synthetic Aperture Radar(SAR)systems and ship target detection algorithms in SAR images.Then,the research history and current status of ship target detection algorithms and information geometry methods in SAR images are discussed,followed by a brief introduction to the basic mathematical knowledge of information geometry,laying the theoretical foundation for the subsequent proposal of target detection algorithms.A ship target detection algorithm based on Finsler metric for SAR images is proposed to address the problem of high false alarm rate in traditional target detection algorithms under complex sea clutter environment.The statistical information of SAR image pixel intensity is represented by Weibull manifold.The Simple Linear Iterative Clustering(SLIC)algorithm is used for superpixel segmentation of SAR ship images.The shape and scale parameters are estimated for each superpixel,and the Finsler metric is constructed based on the estimated parameters.Finally,the maximum inter-class variance method is used to achieve prominent detection of ship targets.Theoretical analysis and comprehensive experimental results demonstrate the robustness and effectiveness of the proposed method.To further improve the detection performance of the existing SAR image ship target detection algorithm,a SAR image ship target detection algorithm based on curvature features is proposed.The algorithm models sea clutter using the gamma distribution function(GDF)and converts the detection problem into an anomaly detection problem in the GDF space.By utilizing the Finsler properties of the GDF space,the curvature features of each GDF are extracted and the detection is implemented using the one-class support vector machine(OC-SVM).Experimental results demonstrate the sensitivity and efficiency of the Finsler geometric features.Finally,a comparative experiment on the GF-3 dataset confirms the superiority of the Finsler curvature method in SAR image ship target detection.