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.
引文
[1]张仁华.实验遥感模型及地面基础[M].北京:科学出版社,1996.
[2]胡晓雯.基于植被指数的绿地信息提取的比较[J].南京信息工程大学学报(自然科学版),2012,4(5):420-425.
[3]和万荣.基于遥感影像的地物信息自动提取技术研究与实现[D].昆明:昆明理工大学,2008.
[4]梁懋钢,刘太喜.海岸带地物反射光谱特性研究[J].海洋科学进展,1990(4):91-95.
[5]赵英时.遥感应用分析原理与方法[M].北京:科学出版社,2003.
[6]雷婉宁,温仲明.基于TM遥感影像的陕北黄土区结构化植被因子指数提取[J].应用生态学报,2009,20(11):2736-2742.
[7]汪清泓.华南地区亚热带树木叶面积指数的高光谱反演研究[J].热带亚热带植物学报,2018,26(4):323-334.
[8]郭凯,孙培新,张锐,等.利用遥感影像软件ENVI提取植被指数[J].新疆环境保护,2005,27(1):10-12.
[9]田庆久.植被被指数研究进展[J].地球科学进展[J]. 1998,13(4):327-333.
[10]陈述彭.地学的探索(第三卷)遥感应用[M].北京:科学出版社,1990.
[11]万余庆,谭克龙,周日平.高光谱遥感应用研究[M].北京:科学出版社,2006.