SAR Ship Detection And Edge Segmentation Based On Multifractal Theory

Synthetic Aperture Radar can monitor the sea surface environment all day and all weather,and has a very broad application prospect.However,in complex natural backgrounds and low SNR scenarios,SAR target detection and edge segmentation become very difficult,and traditional detection algorithms are difficult to meet application requirements.Fractal geometry is very suitable to describe irregular objects,such as complex scenes with nonlinear,non-stationary,and statistical self-similarities.Since SAR sea surface images mostly exhibit non-stationary random characteristics and statistical self-similar features,fractal theory can be well applied to SAR marine target detection.Based on the latest development of fractal theory,this paper focuses on the related theories and methods of SAR ship target detection and edge segmentation in complex sea background,including the following contents.(1)Firstly,the SAR imaging model and parameter technology based on 2D-FBM scene modeling are introduced.Based on this,the terrain modeling and electromagnetic scattering model based on 2D-mFBM are studied.By analyzing the multi-fractal spectrum of simulated SAR imaging,it is found that their SAR images have certain multi-fractal characteristics whether they are HH or VV polarization.Therefore,the feasibility of applying multi-fractal theory to SAR imaging and feature extraction is verified from numerical analysis,which provides some theoretical support for the further application of multi-fractal theory.(2)The paper extends the singularity power spectrum analysis to 2D.By combining 2D-SPS and 2D local singularity index estimation method,a 2D-SPS-based SAR ship target detection method is proposed.And the detection process and specific algorithm implementation process are also given.In the experimental stage,the effective eigenvectors are extracted from the singular power spectrum of SAR images,and then simulated by OpenSAR dataset.Finally,the detection rate is greater than 97% under different false alarm rates.Compared with traditional CFAR detection,our method has obvious advantages.In addition,in order to explore the applicability of the proposed fractal detection operators as much as possible,the simulation experiments are carried out on SAR ship targets with different SNR and different types of SAR ship targets.(3)The paper combines the local singularity index estimation of SAR image with its multifractal spectrum,and proposes a SAR ship target edge segmentation method based on multi-fractal theory.By segmenting the typical real SAR ship targets and comparing with the results of three other traditional edge segmentation algorithms,it shows that the proposed method has better average segmentation performance and can retain more abundant target edge detail features.But at the same time,there is a certain degree of risk of increasing false alarms.The research in this paper enriches the SAR ship detection and edge segmentation methods and techniques,and solves the problem that the traditional algorithm is not ideal in complex natural background and low SNR scenarios.Therefore,it has good theoretical significance and application value.