SPAD Inversion Of Lycium Barbarum Canopy Based On UAV Hyperspectral Image

Lycium barbarum is a geographical indication industry in Ningxia Hui Autonomous Region and a"red card" for Ningxia to go to the world.It is of great strategic significance to the economic development of Ningxia,and the chlorophyll content is an important index to measure the photosynthetic capacity of Lycium barbarum and reflect the growth status of Lycium barbarum.With the development of hyperspectral technology,it is possible to use hyperspectral technology to remotely sense the chlorophyll content of Lycium barbarum.The study area is located in the Lycium barbarum planting base in Tongxin County,Ningxia Hui Autonomous Region.The hyperspectral image of Lycium barbarum canopy in full fruit period is obtained by using the UAV equipped with Hyperspectral camera.The SPAD(Soil and Plant Analyzer Development)value that can reflect the chlorophyll content of Lycium barbarum canopy leaves is synchronously obtained by using a rapid SPAD tester on the ground.The spectral reflectance data.is extracted after geometric correction and radiation correction.Firstly,the spectral reflectance curve is smoothed by Savitzky-Golay filter.After smoothing,the reflectance data R are processed by five spectral transformation methods:first derivative R’,logarithm Log(R),normalization N(R),logarithm derivative Log(R)’and normalized derivative N(R)’.Random leapfrog(RF)and principal component analysis(PCA)algorithms were used for feature extraction of six spectral data,and the vegetation index and spectral parameters based on the original spectral data and the first derivative spectral data were calculated.The correlation analysis was used to screen out the vegetation index and spectral parameters sensitive to the SPAD value.The SPAD value estimation model based on partial least squares(PLSR),support vector machine regression(SVR)and back propagation neural network(BPNN)algorithm was constructed,and the model accuracy based on different spectral transformation methods and different feature extraction algorithms was compared.Finally,a SPAD prediction system based on the optimal feature extraction algorithm and optimal model is designed and made by MATLAB,and the spatial distribution map of SPAD value in the study area is made by ENVI.The results showed that the precision of UAV hyperspectral inversion of SPAD value of Lycium barbarum canopy was high,which provided a new technical means for monitoring the chlorophyll content of Lycium barbarum in the field to guide the fertilization regulation of Lycium barbarum.The main conclusions are as follows:(1)Among the five spectral transformation methods,the first derivative transformation has the most stable effect in improving the correlation between spectral data and SPAD,followed by the first derivative transformation of logarithmic spectrum.(2)The correlation analysis between spectral reflectance data and SPAD value shows that the characteristic wavelengths sensitive to the change of SPAD value are 712nm,716nm and 720nm.(3)Comparing the two feature extraction algorithms,it is found that the modeling effect based on PCA feature band extraction algorithm is better,and the R2 of PCA-PLSR model is 10%higher than that of RF-PLSR model.The R2 of PCA-BP model was increased by 74.4%compared with RF-BP model.(4)Among the 10 spectral parameters,the minimum amplitude λmin showed a high correlation with the SPAD value,and the correlation coefficient reached-0.546;Among the 5 vegetation cover indices,MCARI(Modified chlorophyll absorption vegetation index)was correlated with SPAD value,and the correlation coefficient was 0.455.(5)Among the three established modeling algorithms,the PLSR model has the best prediction effect.Among them,the PLSR model based on the three principal components of the first-order differential spectrum has the highest estimation accuracy with a validation set R2 of 0.877,followed by BP with a validation set R2 of 0.732 and SVR model with a R2 of 0.637.(6)The graphical user interface for predicting the SPAD value of Lycium barbarum canopy was designed by using MATLAB GUI tools,and the spatial distribution map of the canopy SPAD value of Lycium barbarum in the study area was prepared and analyzed.The results showed that the SPAD value of Lycium barbarum in the study area was normal and the Lycium barbarum grew well.