基于高光谱卫星数据的土地沙化识别及提取研究  

The study of land desertification recognition and extraction based on hyperspectral satellite data

作　　者：王建华 李阳[2] 梁树能 孙小飞[4] WANG Jian-hua;LI-yang;LIANG Shu-neng;SUN Xiao-fei(Aerospace Information Research Institude,Chinese Academy of Sciences,Beijing 100094,China;Beihang University,Beijing 100191,China;Land Satellite Remote Sensing Application Center,MNR,Beijing 100048,China;Southwest Jiaotong University,Chengdu 611756,China)

机构地区：[1]中国科学院空天信息创新研究院,北京100094 [2]北京航空航天大学,北京100191 [3]自然资源部国土卫星遥感应用中心,北京100048 [4]西南交通大学,成都611756

出　　处：《华北地质》2022年第4期60-67,共8页North China Geology

基　　金：国防科工局国防科技重点实验室稳定支持专题项目“遥感信息与图像分析技术(JCKY2021201CRS021)”。

摘　　要：研究区位于黄河中游风沙砒砂岩区域,为毛乌素沙漠与黄土丘陵沟壑区的过渡地带,是黄河的粗泥沙集中来源区,也是黄土高原侵蚀较为严重的地区之一。随着遥感技术的发展,高光谱遥感为开展全球性和区域性的土地沙化研究提供了新的手段,土地沙化监测向定量化遥感方向发展。本文利用高光谱数据的精细光谱特征优势,基于资源一号02D卫星高光谱数据反演得到植被覆盖度、裸土指数、地表反照率、改进土壤调整植被指数,将这四种指标作为土地沙化综合监测指标,对研究区的土地沙化程度进行人机交互解译工作,建立了黄河中游典型区域土地沙化遥感解译标志,将人机交互土地沙化监测结果作为参考数据,建立了针对研究区域的决策树分类模型,实现土地沙化的识别、提取及评估,采用目视方法和接收者操作特征曲线对结果进行检验,结果套合情况较好,基于决策树分类模型的方法比人工交互解译方法得到的土地沙化监测结果更加精细化。Study area is located in the sediment-rock region of the middle Yellow River basin,which is the transition zone between the Mu Us sand land and the Hilly Loess region;The coarse sediment of the Yellow River mainly originates from this area which is also one of the areas with serious erosion on the Loess Plateau.With the development of remote sensing technology,hyperspectral remote sensing provides a new method for global and regional land desertification research,and land desertification monitoring is developing towards quantitative remote sensing.This paper uses the advantages of hyperspectral data to retrieve the vegetation coverage,bare soil index,surface albedo and improved soil adjusted vegetation index.Based on the hyperspectral data of ZY-102D satellite,these four indicators are taken as the comprehensive monitoring indicators of land desertification.Through the human-computer interactive interpretation of the land desertification degree in study area.The remote sensing interpretation signs of land desertification in typical areas of the middle Yellow River basin have been established,and the monitoring results have been taken as reference data.The decision tree classification model for the study area has been established to realize the recognition,extraction and evaluation of land desertification.The data have been checked by visual methods and receiver operation characteristic curves,and the results show that there is a significant match,and the method based on decision tree classification model is more detailed than the human-computer interactive interpretation method.

关 键 词：高光谱卫星 土地沙化 黄河中游 决策树 

分 类 号：TP751[自动化与计算机技术—检测技术与自动化装置]