高光谱成像技术实现马铃薯叶片叶绿素无损检测

针对马铃薯叶片，依托高光谱成像技术实现叶片叶绿素含量的无损检测。利用相关性分析获得马铃薯叶片叶绿素敏感波段，结合植被指数，建立基于光谱导数、植被指数的叶绿素含量传统预测模型与贝叶斯正则化-反向传播（BR-BP）神经网络模型。以489 nm光谱一阶导数值、修正型叶绿素吸收植被指数（MCARI）、陆地叶绿素指数（MTCI）为自变量建立BR-BP神经网络模型，其校正集决定系数、预测集决定系数、均方根误差分别为0.8464，0.6804，0.0746。研究表明，传统模型中光谱一阶导数-幂函数模型可较为准确地预测叶绿素含量，BR-BP神经网络模型相比传统预测模型具有更高的预测精度，因此可以实现马铃薯叶片叶绿素无损检测。

Non-destructive detection of potato leaf chlorophyll with hyperspectral imaging technology

Based on the hyperspectral imaging technology, the non-destructive detection of the chlorophyll content of potato leaves is realized. Using correlation analysis to obtain the chlorophyll sensitive band, and combining with the vegetation index, the traditional prediction model of chlorophyll content and the Bayesian regularization-back propagating (BR-BP) neural network model based on spectral derivative and vegetation index were established. The BR-BP neural network model was established with the first derivative value of the 489 nm spectrum, MCARI, and MTCI as independent variables. The coefficients of determination for the calibration and validation sets, and the root mean square error were 0.8464, 0.6804, and 0.0746, respectively. The results showed that the spectral first-order derivative-power function in the traditional model could accurately predict the chlorophyll content. Compared with the traditional prediction model, the BR-BP neural network model had higher prediction accuracy. Thus, it could realize the non-destructive detection of potato leaf chlorophyll.