利用高光谱微分指数反演油茶炭疽病病情指数的研究

Using hyper-spectral derivative indices to inverse Colletotrichum gloeosporioides disease indices

通过实地调查油茶炭疽病病情指数,并使用美国ASD公司生产的手持式野外光谱辐射仪测定相应的油茶冠层光谱反射率,然后将光谱数据进行一阶微分与滑动平均滤波相结合的预处理,经相关分析,选取与病情指数极显著相关的微分指数,采用单变量线性和非线性回归方法,选取部分样本建立油茶炭疽病病情指数的反演模型,并利用其余样本对模型进行精度检验。结果表明,随着病情指数的增大,油茶冠层光谱的一阶微分值在可见光区域的反射峰和吸收谷逐渐消失,红边斜率逐渐减小;病情指数与冠层光谱一阶微分值在480～513 nm、526～569 nm、583～607 nm和669～727 nm 4个波段达到极显著相关;以SDr′为自变量的对数模型反演病情指数的精度最高,其计算出的预测值与实测值之间的相关系数r和均方根误差分别为0.869和0.067,预测精度较高。该研究结果表明利用高光谱微分指数反演油茶炭疽病病情指数是可行的,并为油茶林的健康评价提供了新的方法。

Remote sensing technology has made it possible to monitor vegetation under a range of environmental stress conditions.Several research results have illustrated the superior indicator functions of plant spectral reflectivity derivative over original data.To further investigate the application of remote sensing technology in monitoring Colletotrichum gloeosporioides,this paper used spectral reflectivity of oil camellia canopy measured by hand-held outdoor spectral radiometer（ASD,made in USA） in C.gloeo-sporioides disease index（DI） field survey.The first derivative of hyper-spectral data integrated with a moving average filter was pretreated.Through relevant analysis,the first derivatives highly related with DI were selected.Then using single variable linear and nonlinear regression methods,partial samples were chosen to build an inversion model.Accuracy test was subsequently ac-complished using other tests.The results showed that reflection peaks and valleys of the first derivative of oil camelliae canopies in the visible-light region vanished gradually along with decreasing red-edge slope.A high correlation was noted between DI and the first derivative data in the regions of 480~513 nm,526~569 nm,583~607 nm and 669~727 nm.Using SDr′ as independent variable,the logarithmic model of inversed DI was the most accurate.The correlation coefficient R and RMSE between the predictive and observed values were 0.869 and 0.067,respectively,and also with much higher prediction accuracy.This study showed the feasibil-ity of using the first derivative of hyper-spectral data to inverse C.gloeosporioides DI.This approach could be used to assess the health of oil camellia.