草原火灾一直是影响我国草原生态系统的重要因素, 对其展开研究具有重要的现实意义。 遥感技术的发展, 为草原火灾的相关研究提供了详尽准确的数据, 节省了大量的人力物力和财力。 但在实际应用中, 如何识别不同可燃物及裸土一直是难点问题。 试图将分形理论应用到枯草及裸土的光谱识别研究中, 为上述问题寻找新的研究思路和方法。 该研究运用美国ASD公司的FS3地物光谱仪对吉林省长岭县西部草场的优势种羊草（Leymuschinensis）、 芦苇（Reed）、 虎尾草（Chlorisvirgata）、 全叶马兰（Kalimerisintegrifolia）、 蒙古蒿（Artemisia mongolica）的枯萎植株及裸土进行了野外光谱测量。 并利用Matlab工具, 对上述研究对象的平均野外反射光谱曲线进行了包络线提取和分形盒维数的计算。 经过光谱分析, 裸土和虎尾草的野外光谱反射强, 羊草、 芦苇和蒙古蒿的野外光谱反射相对较弱, 但彼此的谱型较为相近, 较难识别区分。 而全叶马兰的平均野外光谱反射强, 且谱型与其他研究对象有较大差异, 可以很好的识别。 通过分形分析, 研究对象的平均野外反射光谱曲线与包络线具有典型的分形特征。 以分维数作为分类指标, 对研究对象进行聚类分析, 完成了不同地物的野外光谱识别。 该方法与传统光谱分析的方法相比, 能够更好地利用数学方法, 客观的通过识别参数对目标进行识别, 为今后进行其他种类枯草的识别和提取研究提供了新的思路和方法。

Grassland fire disaster is an important influence factor to grassland ecological system in China. Therefore, it is crucial to study on the monitoring, prediction and management of grassland fire. Remote Sensing (RS) provides detailed data and saves a lot of manpower, material resources and financial resources on the research of grassland fire. However, it is difficult to identify the grassland fuel and soil with Remote Sensing. In this paper, we introduced fractal into the spectral analyses of the field-derived spectra (FDS) of grassland fuel and soil to solve the problem above. The study area laid on the Westward of Changling, Jinlin province, China. Study subjects included soil and dominant species: Leymus chinensis, Reed, Chloris virgate, Kalimeris integrifolia and Artemisia mongolica. FDS of study subjects were measured with ASD FS3 and continuums of FDS were calculated by Matlab 2010. Meanwhile, Box-counting values of FDS and continuums were calculated by Matlab 2010. According to the spectral and continuum analysis, it is difficult to identify soil, Leymus chinensis, Reed, Chloris virgate, and Artemisia mongolica because of the similar spectral curves. However, the Artemisia mongolica can be identified for the strong reflection. For typical fractal characteristics of FDS and continuum, clustering analyses of study subjects were done according to box-counting values of FDS and continuum. The results of clustering analyses show that Box-counting values of FDS and continuum are important indexes to identify the study subjects. This study provides a new thought to identity the grassland combustibles and soil with Remote Sensing.