基于多光谱数据的黄河三角洲岸线自动提取  

Study on automatic extraction of coastline in the Yellow River Delta based on multi-spectral data

作　　者：乔学瑾 王庆[1] 战超[1] 王昕[1] 王红艳[1] 杜国云[1] 李雪艳[1] 

机构地区：[1]鲁东大学海岸研究所,山东烟台264025

出　　处：《海洋学报》2016年第7期59-71,共13页

基　　金：国家自然科学基金项目(41271016;41471005)

摘　　要：基于多光谱数据对海岸线自动提取的问题研究已久。针对国内外岸线提取方法较为单一的现状,提出兼顾光谱特征与空间关系的海岸线自动提取方法:将2014年黄河三角洲Landsat8-OLI影像与实测地物反射率光谱对比,选择敏感波段建立提取模型,之后对研究岸段进行分类并自动提取,同时基于908专项山东省修测海岸线标准的目视解译方法对实验部分的海岸线进行提取,最后通过ROC曲线原则对提取结果分别进行0.5与1像元精度评价,实验证明该方法在1个像元精度(提取置信度均高于90%)范围内能快速、准确地提取黄河三角洲复杂地貌类型地区的海岸线,所得到的矢量岸线可直接用于地理信息系统(GIS)分析。The automatic extraction of coastline based on multi-spectral data has been an important question which has received attention for a long time. In this paper, aiming at the situation of the coastline extraction with a rela- tively single method based on either the spectral characteristic or the spatial relation, the authors present an auto- matic extraction method which concludes both spectral characteristics and spatial relations. Firstly, through com- paring the measured spectrums with the LandsatS-OLI image of 2014 and choosing sensitive bands to build extrac- tion model, we classified and extracted the coastline of the Yellow River Delta. Secondly, through using the meth- od of visual interpretation, we extracted the coastline of experimental coastal section based on the coastline revision of Shandong Province in the 908 special project. Thirdly, through using the ROC （Receiver Operating Characteris- tic） curve, we evaluated the accuracy of the extraction result by 0. 5 pixels and 1 pixel individually. Finally, the experimental result of vector-based coastline was directly applied in the subsequent GIS analysis. On the whole, through this new method, we can extract the coastline of Yellow River Delta quickly with an accuracy of one pixel and an extraction confidence level of higher than 90~.

关 键 词：海岸线 自适应阈值 自动提取 精度评价 

分 类 号：P715.7[天文地球—海洋科学]