基于高光谱图像的枸杞叶片叶绿素含量估算模型

Estimation Model of Chlorophyll Content in Lycium barbarum Leaves Based on Hyperspectral Images

为高效无损获得枸杞叶片叶绿素含量,基于无人机搭载高光谱成像仪实拍的宁夏枸杞叶片高光谱图像,结合手持叶绿素仪测定的叶片叶绿素含量,通过对高光谱图像进行特征波段选取并利用光谱微分技术处理,以反射率一阶导数为输入变量,设置不同训练集及数量,选用BP、Elman两种神经网络并分别用遗传算法(GA)、粒子群优化算法(PSO)进行优化,通过预测效果比较确定枸杞叶绿素含量预测的最优模型。结果表明,相比于传统的BP、Elman神经网络,优化后的模型预测结果更好,尤其PSO-Elman模型,预测模型变量为原始反射率与反射率一阶导数的R2分别为0.91408、0.98967。本研究结果可为宁夏枸杞的生产管理提供一定的技术支持。

In order to obtain the chlorophyll content of Lycium barbarum leaves efficiently and nonde-structively,the hyperspectral images of Ningxia Lycium barbarum leaves obtained by unmanned aerial vehicle with hyperspectral imager were used to establish the estimation model combined with the chlorophyll content of leaves measured by hand chlorophyll meter.The characteristic bands of hyperspectral images were selected and processed by the spectral differentiation technique,the first-order derivative of reflectivity was used as the in-put variable,different training sets and numbers were set,the BP and Elman neural networks were selected and optimized by genetic algorithm(GA)and particle swarm optimization(PSO)respectively,and then the optimal prediction model of chlorophyll content of Lycium barbarum was determined after comparing the predic-tion effect.The results showed that compared with the traditional BP and Elman neural networks,the optimized models had better prediction results,especially the PSO-Elman model,whose R2 with the primary reflectance and the first derivative as input variables were 0.91408 and 0.98967,respectively.The results of this study could provide some technical support for the production and management of Lycium barbarum in Ningxia.