Coastline Detection Of SAR Image

SAR image coastline detection shows great value in coastal map drawing,automatic navigation and coastal marine resource management,In recent years,super-pixels had been widely used in image segmentation.The merging model based on super-pixels has also shown great value in extracting the boundary of SAR images.However,different methods of super-pixel segmentation make different results,and so will different merging strategies.Inappropriate combine of the two will be worse.Considering that SAR image segmentation often encounters many problems,such as too much gray level,fuzzy boundary,and speckle noise,it is necessary to use appropriate super-pixel segmentation method and merging model to achieve better segmentation results.By comparing the results of initial segmentation of SAR image obtained by two super-pixel segmentation methods,which are acknowledged to have better edge preservation.It shows that SLIC method has a better segmentation result,it can get a better accuracy in weak boundary area segmentation and a more regular shape,this will be helpful to the later merging.Considering characteristics of SAR images,the land area carries more texture features and information than the sea area.Therefore,the gray-based merging strategy is not enough,so,LBP/LC operator is added to improve this,it will help describe the feature of texture.The texture and gray distances between regions are calculated by G-statistics,in which shape constraint parameters is added.In image preprocessing,considering that speckle noise of SAR image is a multiplicative noise,normal de-noise method and image enhancement method are not suitable for SAR image.After comparing different methods results,Lee filter is chosen to de-noise SAR image,and gray transformation is used to enhance the image.The simulation and experiment proves that after SAR image preprocessing,the coastline segmentation of SAR image can obtain a better segmentation result.Compared with the traditional method,the proposed method not only extracted the coastline of SAR image correctly,but also improved the accuracy.In the term of time efficiency,it also has a better performance.Therefore,the method proposed by this paper is valuable in coastline segmentation.