基于高光谱数据的杉木和马尾松识别研究

Discrimination of Cunninghamia lanceolata and Pinus massoniana based on hyperspectral data

高光谱具有波段窄、波段多的特点,能够提供比多光谱遥感更精细的地物光谱信息,为识别光谱性质相似的森林树种提供了有效途径。为了识别光谱性质相似的杉木和马尾松,利用逐步判别分析法对杉木和马尾松的原始光谱、一阶微分光谱、对数变换后取一阶微分光谱及植被指数进行分析,结果表明:对数变换后取一阶微分的光谱及植被指数对杉木和马尾松的识别精度最高,分别为96.67%、89.17%;两树种识别的特征波段为490～499nm,500～519nm,560～579nm,610～619nm,630～649nm,680～700nm,710～719nm,770～779nm;组成植被指数NPCI、mND705、SRPI、GNDVI、GM的波段为:430、445、550、680、700、705、750nm。

Hyperspectrum provides an efficient method for the discrimination of forest trees with similar spectrum, because hyperspectrum has the characteristic of narrow band and multiband, and can provide more micromesh spectral information than multispectrum. To discriminate Cunningharnia lanceolata and Pinus massoniana, the original spectrum, first derivative of spectra, first derivative of log（R）, vegetation indices of the two trees were analyzed by the method of stepwise discriminate analysis. The results show that： first derivative of log（R） and vegetation indices were efficient for the discrimination of these two trees, and the accuracies reached 96.67% and 89. 17%0, re spectively. The characteristic bands for discriminating the two trees were 490-499 nm, 500-519 nm, 560-579 nm, 610-619 nm, 630-649 nm, 680-700 nm, 710-719 nm, 770-779 nm. The bands of 430, 445, 550, 680, 700, 705, 750 nm composed the vegetation indices for the discrimination of the two conifer species.