Fast SAR Image Segmentation Based On Merging Measure And K-S Test

Synthetic aperture radar(SAR),is a tool with high resolution and ability to acquire and detect important information resources.With the extensive application of synthetic aperture radar,synthetic aperture radar has high resolution in the case of all-weather,all day long time to work,and is widely used in military reconnaissance,ocean surveillance,the field of natural disaster detection.SAR image is the direct information provided by synthetic aperture radar,and in many cases,we tend to be interested in some of the targets,so we need SAR image segmentation technology to solve this problem.Because of the speckle noise of SAR image,the quality of SAR image segmentation is affected,and the difficulty increases.In recent years,many researchers have proposed a large number of SAR image segmentation algorithms,these methods can be divided into two categories: SAR image segmentation algorithm based on image features and SAR image segmentation algorithm based on model optimization.There are three kinds of SAR image segmentation algorithm of image features: SAR edge information of the image segmentation algorithm,SAR image region information segmentation algorithm and segmentation algorithm of SAR image edge information and region information based on mixed.And there are four main methods of SAR image segmentation based on Model Optimization: variational method,the segmentation method based on the shortest description length criterion,Markov random field method and graph theory method.These methods have their unique advantages in the segmentation of SAR images,but sometimes it is not ideal because of the high complexity,especially with large area in the region merging process.This paper focuses on the SAR information after data processing,in order to solve the problem of traditional SAR image target is not clear to the naked eye is difficult to distinguish the target and area,first on SAR image by weighted median filtering pretreatment Gauss smooth,and then use the water ridge transform based on RESM to get the initial segmentation results of SAR images.Finally,based on the relative public boundary length and K-S combined with the distance measure for SAR image segmentation.This method can solve the problem that the image processing is slow and the quality is not ideal in the traditional method.It can be quickly segmented and can adapt to most SAR image data at the same time.The main work of this paper includes the following aspects:1.According to the characteristics that the SAR image is easy to be affected by the strong scattering point of the scene,the weighted median filter is used to preprocess the SAR image.In this paper,the Gauss function was used to smooth the weighted window,and the influence of weighted windows with different sizes on the processing results was discussed.2.This paper explained the initial segmentation method used in SAR image segmentation.and used the watershed transform based on RESM to realize the initial segmentation of SAR image,and introduced the initial segmentation method based on SRG.Then,the simulation results were compared.3.The final segmentation of SAR image required the region merging of the initial segmentation results.Two merging measures were proposed,including merging measures based on relative common boundary length and merging measures based on K-S distance.In this paper,by setting the threshold in the area was small by using the relative public boundary length merger measure based on merger,measure based on K-S distance in using large area,and finally using RAG and NNG methods to get the final segmentation results,and through simulation experiments based on the relative public boundary length and statistical similarity measure method in this paper,and the segmentation results were compared.At the same time,we combined the third chapter and the fourth chapter,and obtained the final segmentation results,and compared the operation time.