基于知识的多时相TM图像森林火烧迹地快速提取方法

Method of Fast Extracting Fire Scars from Multi-temporal TM Images on the Basis of Knowledge Available

本文以覆盖福州市及周边地区的五期TM图像为例,进行森林火烧迹地的自动提取研究。分别对TM图像上火烧迹地及其背景地物采样,分析火烧迹地在TM图像6个波段及PC变换、TC变换、多时相植被指数(MTVI)上的光谱特征,发现了火烧迹地与背景地物区分的光谱知识。其次,利用DEM数据分析火烧迹地在地形上的分布特点,发现了火烧迹地空间分布的知识规则。在此基础上建立了基于知识的多时相TM图像上森林火烧迹地提取模型,从而解决了快速识别和提取森林火灾信息的问题。

<Abstrcat> Multi-temporal TM images covering Fuzhou city and its neighboring area are selected as case study in order to extract fire scars quickly. Spectrum characteristics of fire scars and background surface features among six bands of TM images, principal components (PC) change, Tasseled Cap (TC) change and multi-temporal Vegetation indices (MTVI) are analyzed. Spectrum knowledge to differentiate fire scar from background surface features is discovered. Spatial distribution regularity of fire scars on the topography is investigated by use of DEM data. Then a model for forest fire scar extraction on Multi-temporal TM images is developed. So, forest fire scars can be identified and extracted quickly.