Automatic extraction of coastline by remote sensing technology based on SVM and auto-selection of training samples

Abstract

The timely and accurate automatic extraction of coastline from satellite remote sensing imagery is one of the important applications of remote sensing technology and has great significance for management planning of the sea area.Because the spectral characteristics of coastal water are susceptible to regional environment,the traditional method of normalized difference water index(NDWI)threshold segmentation may easily misclassify water as land in the process of separation of land and water,which will seriously affect the accuracy of shoreline extraction.In this paper,on the basis of NDWI model,the authors proposed an automatic coastline extraction method based on classification sample auto-selection and support vector machine(SVM).Firstly,through the NDWI calculation and global threshold segmentation,the initial water distribution information is obtained.And then,the classification samples are selected automatically under the control of NDWI information.Thirdly,the water are separated from the land by using SVM classifier.The last step is to fill small terrestrial water body units and track coastline automatically.The experimental results show that this method can effectively enhance the capability of coastal water identification and improve the accuracy and automation of the coastline extraction from remote sensing imager.