Research Of High Resolution SAR Image Water Extraction Algorithms

SAR image reflects the geographical characteristics of microwave scattering,water area of high resolution SAR images belong to a smooth surface,it reflects most of the microwave,and it will absorb the microwaves emitted radar systems,and thus part of the receiving water body will be relatively weak scattering signals,resulting in SAR image corresponding to the low performance of gray on gray image.Since the SAR image imaging modality for the slant range imaging,in many mountainous topography obvious SAR image appears shaded geometric deformation phenomena.These shadows will seriously affect the water body recognized image information.This paper is to study the high-resolution SAR images of water extraction algorithm,the main water extraction algorithm for SAR image divided into three categories: threshold,curve evolution method and clustering method.This paper mainly on the first two methods analysis and research to identify the deficiencies,to improve it or make a new algorithm.Based on high resolution SAR image imaging characteristics,the purpose of this paper is to improve SAR image of water extraction algorithm.Since the SAR image imaging modality for the slant range imaging,in many mountainous topography obvious SAR image appears shaded geometric deformation phenomena.These shadows will seriously affect the water body recognized image information.According to the characteristics of SAR image body shadow,this paper presents a fast,accurate,automatic removal of water extraction to cull hillshade.The method uses Otsu segmentation method DEM data,elevation data and the value of large water bodies crude extract data matching,so as to determine and remove the shadow part of the mountain.Extraction results show that this method can improve the accuracy of the extraction of water,reduce water extract false alarm rate.Using the level set method for water SAR image extraction process,kernel functions are used to evolve the level set,they can better express the same gray scale information area,while a certain extent in SAR image noise,increased step long,reducing the level set iterations,accelerate the speed of evolution,at the same time,it can be able to obtain more segmentation results.