高分一号遥感卫星数据林地信息快速提取方法的研究  

Study on GF-1 Remote Sensing Satellite Data Forestry Information Fast Extraction

作　　者：周日平 Zhou Riping(Dadi High-tech Geological Exploration Branch, China Coal Geology Group Co. Ltd., Beijing 100075)

机构地区：[1]中煤地质集团大地高科,北京100075

出　　处：《中国煤炭地质》2019年第2期72-76,共5页Coal Geology of China

基　　金：全国自然资源遥感综合调查与信息系统建设

摘　　要：不同类型的林地具有不同的反射率特征,由此计算获得林地识别与分类的归一化植被指数(NDVI)。NDVI可参与对高分一号遥感图像的监督分类:首先根据图像光谱特征及NDVI参数,在ENVI专业图像处理平台上,选择各类林地的典型样本区,进行计算机自动信息提取,获得每个像素点属于某类林地的相似度信息,然后结合纹理、形状、空间关系等特征,利用聚类分析和迭代运算等综合处理方法,快速得到具有实用性的专题分类矢量结果。该方法光照度相对较弱的阴坡和薄云覆盖下的林地分类识别效果也比较好。通过对光谱角填图法、马氏最小距离法、最大似然法等监督分类的结果进行比较,认为采用光谱角填图法提取高分一号卫星原始数据的林地矢量信息精度最高。Different forestry types have different reflectance features, the normalized difference vegetation index (NDVI) computed from the features can participate GF-1 remote sensing images supervised classification. Firstly, according to image spectral features and NDVI parameters select a variety of forestry typical sample areas from the ENVI professional image processing platform carry out computer automatic information extraction to obtain similarity information of every pixel belongs to a kind of forestry. And then, combine with features of lamination, shape and spatial relationship etc. through synthesized clustering analysis and iterated operation, fast obtain practical thematic classification vector results. The method has better effect of classified identification for forestry in shady slope with relatively weak light or covered by thin cloud. Through comparison of supervised classification results from the spectral angle mapping method, Markov minimum distance method and maximum likelihood method, have considered that the forestry vector information accuracy from spectral angle mapping method extracted GF-1 satellite primary data is the highest.

关 键 词：植被指数 波谱特征 林地信息 光谱角填图法 

分 类 号：P627[天文地球—地质矿产勘探]