面向对象的极高海拔区水体及冰川信息提取——以珠穆朗玛峰国家级自然保护区核心区为例  

Object-oriented Information Extraction of Water Bodies and Glaciers in Extreme High Altitude Area:A Case Study of the Core Area of Mt.Qomolangma (Everest) National Nature Preserve

作　　者：张继平[1,2] 刘林山[1] 张镱锂[1] 聂勇[1,2] 张学儒[1,2] 张琴琴[1,2] 

机构地区：[1]中国科学院地理科学与资源研究所,北京100101 [2]中国科学院研究生院,北京100049

出　　处：《地球信息科学学报》2010年第4期517-523,共7页Journal of Geo-information Science

基　　金：中国科学院对外合作重点项目(GJHZ0954);国家重点基础研究发展计划项目(2005CB422006);HKKH生态系统管理合作项目

摘　　要：极高海拔地区多为河流发源、冰川发育地,由于地形起伏强烈,且野外考察验证工作困难,传统的遥感信息提取方法很难保证该地区水体及冰川的提取精度。本文基于ASTER影像,运用面向对象的图像信息自动分析方法,对珠穆朗玛峰国家级自然保护区核心区的水体及冰川信息进行了提取研究。为保证信息提取的准确度,将数字高程模型(DEM)及其衍生数据(坡度、坡向),归一化植被指数(NDVI)数据,及有助于区分水体、冰川与其他地物的相关指数(冰雪指数NDSII)及波段运算结果(b1-b3)、(b3/b4)等,分别作为一个波段叠加到原始图像中,使之成为对目标地物光谱特征的有益补充。并对不同类型的水体及冰川进行多级、多尺度分割,以满足其对分割尺度的不同要求。分割完成后,综合考虑目标地物的光谱特征、纹理特征、空间结构特征,根据各特征指数的直方图信息,设定合适的阈值,建立了各水体及冰川类型信息提取的知识规则,并结合实地调查对信息提取的精度进行验证,改进了ASTER遥感影像自动快速提取极高海拔区水体及冰川信息的实用模型。Extreme high altitude area is the birthplace of rivers and glaciers.It is quite difficult to accurately extract information of rivers and glaciers from remote sensing images due to the strong undulating topography and arduous field verification.This study chose ASTER image as the data source to extract water body and glacier information in the core area of Mt.Qomolangma（Everest） National Nature Preserve（QNNP） by the method of object-oriented automatic classification.In order to ensure the accuracy of information extraction,the digital elevation model（DEM） and its derived data（slope,aspect）,normalized difference vegetation index（NDVI）,as well as other indexes that help to distinguish water bodies and glaciers from other land objects,such as Normalized Difference Snow/Ice Index（NDSII） and band math（b1-b3;b3/b4） were stacked respectively as added spectral bands,which contributed to the supplementation of the spectral characteristics of target surface objects.Then,multi-level image segmentations were carried out for each type of water body and glacier to meet their different requirements of segmentation scale.Image segmentation is the key step of object-oriented automatic classification,which requires many times of repeated experiments to find out the most suitable segmentation scale.After the accomplishment of image segmentation,the spectral characteristics,texture features and spatial structure characteristics of each target surface object were comprehensively analyzed.And then,the knowledge-based extraction principles for each type of water body and glacier were established by determining the thresholds of each index according to their histogram features.The classification accuracy was assessed based on field survey results.According to the accuracy assessment,the total accuracy of image classification is 95.14% which satisfies the accuracy requirement of object-oriented automatic classification.This study improved the model for automatic and quick extraction of water bodies and glacier

关 键 词：面向对象 水体 冰川 信息提取 珠穆朗玛峰国家级自然保护区 

分 类 号：P343.6[天文地球—水文科学]