分形理论在马尾松松材线虫病发病早期高光谱探测中的应用

Application of Fractal Theory in Hyperspectral Detecting the Early Stage of Pine Wood Nematode Disease(Bursaphelenchus xylophilus) of Pinus massoniana with Hyperspectrum

采用便携式ASD野外光谱仪,对3个样地未知病害马尾松进行反射光谱连续测量,在对随机抽取的样木反射光谱数据红边位置、绿峰反射高度和红谷吸收深度分析的基础上,进一步利用分形理论,计算450～780nm之间的反射光谱曲线的分形维数,并分析曲线构形变化机理与分形维数之间的关系。结合后期补充调查,结果表明:(1)分形维数在调查期间特别是后期,呈下降趋势,这基本与3个光谱参数的变化吻合,但相对红边位置、绿峰反射高度和红谷吸收深度,分形维数变化趋势较稳定,说明分形维数作为马尾松健康状况预测的指标具有一定潜力;(2)分形维数与绿峰反射高度和红谷吸收深度存在正相关关系,且与绿峰反射高度的相关关系高于红谷吸收深度;(3)补充调查死亡样木分形维数与绿峰反射高度的关系模型,进一步说明分形维数对马尾松健康状况的分析是可行的。

A portable Spectral Device was used to measure the hyperspectrum of Pinus massoniana in 3 plots for detecting the early stage of Bursaphelenchus xylophilus. We measured the red edge position, green peak height and absorb depth of red band of the sampled trees at random in the 3 plots; and then the fractal theory was used to analyze the spectral curves （450 nm to 780 nm）; finally, we investigated relationships between fractal dimension and the mechanism of spectral curves variation in integration with a post-measurement supplementary survey. The results showed that （1） fractal dimensions had a decreasing trend in the period of investigation, which was basically coincident with the variation of the three spectral parameters, but the change trend of fractal dimensions were more stable than the three spectral parameters; （2） fractal dimensions were positively correlated with green peak height and absorb depth of red band, which could serve as a comprehensive parameter for detecting the early stage of B. xylophilus; （3） the supplementary investigation further supported the relationship model between fractal dimension and green peak height. It is concluded that the fractal analysis of hyperspectal data is feasible for us to describe the change of spectral curves and detect the B. xylophilus.