基于高光谱的鲜桃叶片叶绿素含量检测

Detection of chlorophyll content of peach leaves based on hyperspectral technology

为实现鲜桃叶片叶绿素含量的快速无损检测,使用鲜桃四个生长时期的叶片光谱数据及其叶绿素含量数据,利用原始光谱及其变换形式,采用主成分分析和小波去噪预处理数据作为输入矢量,采用支持向量回归机和偏最小二乘法分别构建基于主成分分析-支持向量回归和小波去噪-偏最小二乘回归两种方法的预测模型,并与传统方法建立的模型结果进行比较。通过实验发现,整体建模结果最优的全生长期数据校正集和验证集模型的R2为0.872 7和0.871 4,RMSE分别为0.156 3和0.154 4;采用传统建模方法时,效果最优的是主成分回归模型,全生长期验证集模型R2为0.825 9,RMSE为0.174。结果表明:采用主成分分析-支持向量回归和小波去噪-偏最小二乘回归建模方法的建模效果均优于传统方法,能够应用于基于高光谱的鲜桃树叶绿素含量检测。

In order to realized fast nondestructive measurement of peach Leaves chlorophyll content, the hyperspectral data and the chlorophyll content of leaves on four growth periods of peach are used as basic data, and the principal component analysis and wavelet denoising preprocessing data of the original hyperspectral data and its transformation data are used as input vectors data. Building respectively the prediction model is based on the two methods of principle component analysis support vector regression （PCA SVR） and wavelet denoising partial least square regression （WD-PLSR）, compared with the model results of traditional methods to establish. Experiments find that the best coefficient of determination （R2） of the calibration set and a validation set of the entire growth period are up to 0. 872 7 and 0. 871 4, the root-mean square error （RMSE） is 0. 156 3 and 0. 154 4 respectively. The model of the optimal is the principal component regression model when building model uses traditional modeling methods. And its coefficient of determination （R2） of the validation set of the entire growth period is 0. 825 9, and the root mean square error （RMSE） is 0. 174 0. The result shows that the model is better than the traditional method to use the modeling method of principle component analysis support vector regression （PCA-SVR） and wavelet denoising partial least square regression （WD-PLSR）. The modeling method can be applied to the detection of chlorophyll content of peach leaves based on hyperspectral technology.