无人机可见光谱识别越冬期油菜叶片叶绿素含量估测研究

Identification of Chlorophyll Content in Leaves of Rapeseed in Overwintering Period by UAV Visible Spectrum

本研究利用湖南农业大学耘园基地60个不同种类油菜越冬期叶绿素含量及叶片SPAD值,结合无人机可见光谱数据进行相关分析。利用Matlab 2020a图像处理系统软件进行图像分析,获取图像中的各项颜色特征,筛选出B、G/R、G/B、G/(R+B)、(G-R)/B、(G-B)/R、(G-R)/(G+R)、(G-B)/(G+B)、(G-R)/(R+G+B)和(G-B)/(R+G+B)等10个与油菜SPAD值显著相关的颜色特征,采用BP神经网络(BPNN)、多元逐步回归(MSR)和多元线性回归(MLR)方法分别构建油菜SPAD值分析模型。结果表明,3种分析方法中,BP神经网络模型精度最高,其模型的R^(2)、RMSE分别为0.461、2.147,模型验证的R^(2)、RMSE、RPD分别为0.367、2.012、1.642。综合分析,无人机可见光谱图像结合BP神经网络模型可以监测不同类型油菜越冬期叶绿素的含量及油菜田间长势。本研究可为降低油菜田间调查成本和促进油菜效益提高及规模化生产提供参考。

In this study,the correlation analysis was carried out by using the chlorophyll content and leaf SPAD value of 60 different kinds of rape in the base of Hunan Agricultural University,combined with the UAV visible spectrum data.The co-lor features in the image were obtained by image analysis using Matlab2020a image processing system software,and 10 color features which were significantly related to rape SPAD were screened out,such as G/R,G/B,G/(R+B),(G-R)/B,(G-B)/R,(G-R)/(G+R),(G-B)/(G+B),(G-R)/(R+G+B)and(G-B)/(R+G+B).BP neural network(BPNN),multiple stepwise regression(MSR)and multiple linear regression(MLR)were used to construct rape SPAD value analysis models.The results showed that among the three analysis methods,the accuracy of BP neural network model was the highest,the R^(2)andRMSE of the model were 0.461 and 2.147 respectively,and the R^(2),RMSE andRPD of the model were 0.367,2.012 and 1.642 respectively.According to the comprehensive analysis,the unmanned visible spectrum image combined with BP neural network model can monitor the chlorophyll content and field growth of different types of rape during overwintering.This study can provide reference for reducing the cost of rape field investigation and promoting rape efficiency and large-scale production.