基于可见-近红外光谱技术的炭疽病侵染后油茶叶片叶绿素含量预测研究

分析炭疽病侵染后油茶冠层的可见-近红外光谱特征,探索建立病害胁迫下油茶冠层叶片叶绿素含量的预测模型。通过实地调查病情指数,获取不同病害程度的油茶冠层叶片光谱数据及其叶绿素含量,并对光谱数据进行了一阶微分与滑动平均滤波相结合的预处理,再通过光谱数据重采样,提取敏感波段建立了叶绿素含量的BP神经网络预测模型。结果表明:(1)随着病情的加重,油茶冠层光谱可见光区域的反射峰和吸收谷逐渐消失;红光到近红外陡峭的红边被逐渐拉平;在近红外区域,健康油茶的光谱反射率明显大于感病油茶的光谱反射率。(2)微分光谱484～512,533～565,586～606和672～724nm四个波段是叶绿素吸收和反射的敏感波段。(3)以敏感波段为输入变量建立的BP神经网络模型,其计算出的预测值与观测值之间的相关系数r和均方根误差分别为0.992 1和0.045 8。因此,利用可见-近红外光谱技术预测炭疽病侵染后油茶叶片叶绿素含量是可行的。

Prediction of chlorophyll content of leaves of oil camelliae after being infected with anthracnose based on Vis/NIR spectroscopy

The prediction model of chlorophyll content of leaves in canopies of oil camelliae under disease was explored and built by analyzing the Vis/NIR spectroscopy characteristics of oil camelliae canopies after being injected with anthracnose. Through field survey of disease index (DI), chlorophyll content and spectral data of leaves in canopies surviving different severity of disease were acquired. The first order differential of spectral data combined with moving average filter was pretreated. The prediction model of BP neural network of chlorophyll content was built by extracting sensitive wave band from spectral resample data. The results showed that with the disease being aggravated, reflection peaks and valleys of spectra of oil camelliae canopies in visible-light region vanished gradually, steep red edges from red light to near infrared leveled little by little, and reflectivity of healthy oil camelliae was far larger than that of ill ones. The sensitive wave band of absorption and reflection of chlorophyll lay in the region of 84-512, 533-565, 586-606 and 672-724 nm. The correlation coefficient r and RMSE between predictive values calculated from BP neural network using sensitive wave band as input variables and observed values was 0.9921 and 0.0458 respectively. It was therefore feasible to utilize Vis/NIR spectroscopy technology to forecast the chlorophyll content of oil camelliae after being infected with anthracnose.