绿洲景观多季相特征的线性光谱混合分解方法研究  

Extracting oasis landscape multi-seasonal characteristics with linear spectral mixture analysis

作　　者：刘娜[1] 李红[2] 孙丹峰[1] 周连第[2] 

机构地区：[1]中国农业大学资源与环境学院,北京100193 [2]北京市农林科学院综合发展研究所,北京100097

出　　处：《农业工程学报》2010年第8期174-182,F0004,共10页Transactions of the Chinese Society of Agricultural Engineering

基　　金：国家科技支撑项目(2006BAB15B05);国土资源部土地利用重点实验室开放研究项目(08-02)

摘　　要：因干旱、半干旱区生态环境的脆弱性和社会经济的干扰,区域景观季相间具有显著变异,但该类区域季相景观异质性研究相对薄弱。加上遥感数据空间分辨率的约束,表征荒漠化状态的异质性景观特征以一定程度的混合像元存在,传统的景观信息提取方法具有一定的局限性。以甘肃省民勤县为例,进行了荒漠化异质性季相景观特征的线性光谱混合分解方法研究,包括端元类型确定,多季相代表性端元选择和光谱值优化等关键步骤。结果表明,多季相线性光谱混合分解模型考虑了端元在季相之间的可比性和分解结果的物理意义,可以直接进行物理解译,有效模拟民勤绿洲的景观要素分布,多季相端元丰度可以描述各类端元的季相变异,为土地荒漠化监测评价提供保障。The frangibility of eco-environment and interference of socio-economy in arid and semi-arid areas often causes high tendency to land degradation and distinct variations in regional landscape between seasons.Moreover, constrained to the spatial resolution of remotely sensed data,high spatial heterogeneity of landscape indicating desertification status often exists in the form of mixture pixels with remote sensing assessment.Hence,traditional methods including pixel-based classification have their limitations.This paper taking Minqin as an example expanded the linear mixture analysis based on multi-seasonal data for extracting seasonal characteristics of desertification heterogeneity including determination of types of end-members,selection of multi-seasonal representative end-members and optimization of end-member spectra.The results indicated that the multi-seasonal linear mixture model allowing for the comparability between seasons and physical meaning of decomposition results could be interpreted physically and effectively simulate the distribution of landscape elements in Minqin oasis and the estimated end-member fractions of seasons could describe seasonal variations of oasis landscape,which could offer support for land desertification dynamic assessment.

关 键 词：主成分分析 遥感 图像处理 线性光谱混合分解模型 多季相景观特征 端元类型 光谱值优化 民勤绿洲 

分 类 号：TP79[自动化与计算机技术—检测技术与自动化装置]