基于统计判别法的不同类型植被高光谱数据分析研究

Research and Analysis on Hyperspectral Sampled Data of Different Types of Vegetation Based on Statistic Discriminance

该文利用统计学方法对不同植物的多个采样光谱样本进行比较分析。选用大豆、红薯、茶树、松树的19个野外采样光谱样本,采用Mann-WhitneyU检验方法和马氏距离判别法进行比较分析,结果显示,Mann-WhitneyU检验适用于同一地物反射光谱的采样样本,判别分析方法适用于不同地物的采样样本,两种方法基本能达到区分的效果。

It is well known that spectral signatures are not unique to plant types.In this paper,19 spectral samples of field from 4 different plants including tea,bean,pine and pachyrhizus are obtained.At first these samples are analyzed by using of Mann-Whitney U-test method,and the result shows that only 8 pairs of spectral samples of same plant are discriminated correctly at the level(α=0.05) and the correct rate is 20 percent.And the obscure samples have 18 pairs and the incorrect rate is 45 percent.Analyzed by using of Mahalannobis distance method,the result is acceptable that only one spectral sample of tea is divided to pine by mistakes.The result shows that Mann-Whitney U-test method is fit for different samples of the same plant,and Mahalannobis distance method is fit for different plants.