Research On SAR Image Segmentation Algorithm Based On G~0 Distribution

Synthetic Aperture Radar(SAR),a coherent wave imaging radar operating in the microwave band,is a kind of active microwave sensor.Not affected by environment,climate and temporal factor,it has superior penetrating power.It is widely used in military exploration,marine monitoring and geological exploration.In the present era,with the development of system technology,more and more SAR images are available,and the interpretation of SAR images becomes very important.SAR image segmentation is an important basis and premise for SAR image interpretation.The segmentation accuracy of SAR images directly determines the quality of SAR image interpretation.Noting that,the coherent wave imaging mechanism of SAR system makes speckle noise become an inherent characteristic of SAR image.The existence of speckle noise makes SAR image segmentation more difficult.Therefore,SAR image segmentation has always been a hot issue in SAR image research.In order to accurately model SAR images and fit the statistical distribution SAR images with small sample,on the basis of comprehensive analysis of the characteristics of the G_I~0probabilistic statistical distribution model,we did some algorithm research on SAR image segmentation based on the G_I~0distribution.Specific research contents are as follows:1)SAR Image Segmentation algorithm Based on Random Weighting Estimators and the Improved Threshold Level Set.In the first place,the G_I~0 was selected the describing of SAR images,and the Random Weighing Estimators(RWE),which solves the Problem of estimation accuracy of the G_I~0 distribution Parameters with small sample,was used to estimate the roughness parameteraand scale parametergof the G_I~0 distribution.Afterwards,we substituted the estimated roughness parameters and scale parameters into the Renyi entropy formula of G~0distribution to obtain the entropy matrix.Finally,the improved threshold-based level set(ITLS)was utilized to achieve SAR image segmentation.The experimental results of synthetic and real SAR images confirm that our segmentation algorithm has higher segmentation accuracy and high stability.2)SAR Image Segmentation algorithm Based on SLIC Superpixel Modeling.In order to reduce the computational complexity,for the selection method of image regions to model,we used SLIC superpixels as the basic unit of modeling.In addition,we carried out the G_I~0distribution modeling for SAR images by utilizing random weighted estimator.Then we achieved multi-region SAR image segmentation based on SAR image segmentation method proposed in 1).The experiments prove that the method does improve the segmentation efficiency on the basis of ensuring segmentation accuracy.And the good segmentation results are achieved for multi-region SAR images.