资源3号卫星影像的植被信息提取

Extraction of vegetation information based on resources satellite No.3(ZY-3) data

资源3号卫星(ZY-3)是我国2012年发射成功的第一颗自主民用高分辨率立体测绘卫星。在对ZY-3卫星数据精细处理基础上,分析了典型地物的光谱特征和纹理特征,构建了适合植被信息提取的ZY-3决策树模型,将研究区地类分为针叶林、阔叶林、灌木林、竹林、耕地、建设用地、裸露地、水体和其他地类共9种地类,开展植被信息提取和精度检验。结果表明:(1)ZY-3比较适合于森林植被信息的提取,特别是针叶林和阔叶林;(2)ZY-3数据的总体分类效果较好,精度达到了91.96%,Kappa系数0.902 9,说明根据影像的光谱特征和纹理特征,采用决策树算法选择合适的阈值提取植被信息是可行的。

ZY-3 is the first high resolution satellite of our country, which was launched in 2012, was used in stereo-survey and drawing including forest and vegetation and was one self-developed and for civilian. Based on the fine-processing of the ZY-3 data of Lijiang River Basin, the spectral features and texture features of typical ground objects in the area were analyzed, ZY-3 decision tree models were established that were suitable for vegetation information extraction, and then the land type if the studied area were divided into 9 kinds, they are needle-leaved forest, broad-leaved forest, shrub, bamboo, farm, buildings, naked land, water and others, finally the vegetation information of all kinds were extracted and the extraction accuracy were tested. The results show that ： （1） ZY-3 is very suitable for vegetation extraction, better for the information extraction of needle-leaved and broad-leaved forest than others; （2） All the land types＇ information extraction from the data are good, the total classification accuracy was up to 91.96%, and the Kappa coefficient was 0.902 9. It is feasible that the studied area vegetation information is extracted according to the spectral features and texture features of the studied area data, by using decision tree algorithm models and choosing an anorooriate threshold value.