MS-CV Method For Extracting Coastline From MS-CV Images

The coastline,the boundary between the sea and the land,is the benchmark for the division of national territory and the exclusive economic zone of the ocean,which is an indispensable important terrain element in charts and topographic maps.Coastline area is the most sensitive area,because of the advantages geographical and frequent human activities.Therefore,it has become a hot issue in the ocean research to extract coastline information quickly and accurately.Synthetic aperture radar(SAR)is very suitable for coastline information detection because of its high frequency,high precision,rapid,unified,synchronous,comprehensive data acquisition capability,the characteristics of all-weather,large area,periodic observation,the ability to penetrate the clouds,mist and sleet.In a word,it has unparalleled advantages in coastal resources survey and coastal zone management.The main body of this paper is how to extract coastline automatically,quickly and accurately for SAR images.On the basis of the existing extraction methods,this paper focuses on the research of the level set C-V model segmentation method,which has strong detection sensitivity and anti-noise ability.Aiming at the problem of iteration complexity and slow boundary detection efficiency in C-V model,which is caused by the speckle noise inherent in the SAR image,the uncertainty of the reflection information in the land and sea area,and the difficult to distinguish the complex ground objects,this paper improves from two aspects:vertical image sequence generation and transverse fine automatic segmentation.Based on the existing level set C-V model for coastline extraction,a new method--Multiscale Sequence-Chan Vese(MS-CV)model is proposed.The main work and innovation of this paper are as follows:1.The multiscale Pyramid image generation method is improved.The characteristics of the coastline segmentation are given by two multiscale image sequence generation based on the exponential and hyperbolic types.The purpose of improving sequence generation is to reduce computation effectively,noise interference and integrate the objects concerned by scaling.2.A fast and automatic segmentation method(MS-CV method)is presented for the SAR image's coastline,and this method changes inheritance iterative way from small scale to large scale image and utilizes auxiliary method,including Butterworth low-pass filtering,Ostu pre-segmentation and comprehensive threshold.It helps to enhance the initial value of C-V model and automatic segmentation of coastline.Experiments show that this method improves the extraction precision and calculation rate of coastline for SAR image.3.A method for evaluating the result accuracy of the coastline is developed,the offset error,that is,the root mean square error of the offset.The difference of coastline between MS-CV segmentation method and conventional multi-scale C-V model is evaluated to solve the problem that the original data points or positions of different methods are different and the accuracy of the coastline cannot be evaluated.In this paper,using Cosmo-SkyMed image data for research experiments,compared with the existing multiscale C-V model segmentation method,the MS-CV method has obvious advantages in the accuracy and calculation rate of coastline extraction.