Inversion And Modeling Of Carotenoid Content Of Dicotyledonous Plant Leaves Based On The Hyperspectral Data

The photosynthesis of vegetation plays a crucial role in the circulation of material and energy of the whole ecosystem.In the photosynthesis,photosynthetic pigments involve in absorption,transfer of light and photochemical reactions,mainly consisting of chlorophyll and carotenoids.Among them,carotenoids play an important role in protecting chlorophyll,maintaining photosynthesis,delaying vegetation leaf senescence,and responding to external stimulus in time and responding,which are of great significance in predicting vegetation health,nutrition and growth status.Estimating its contents accurately had great significance for predicting the health status of vegetation.This research was based on LOPEX'93 database.The system analyzed the respective quantitative relationship between carotenoid content in the leaves of dicotyledonous plants,normalized difference vegetation index(NDVI),the ratio of vegetation index(RVI),difference vegetation index(DVI),re-normalized vegetation index(RDVI),and the principal components of the original spectrum.NDVI,RVI,DVI and RDVI were constructed by any combination of two spectral bands in400-2500nm.It was founded that the correlation coefficients of the four vegetation indices in the optimal band were all greater than 0.88,reaching a strong correlation level.The results of correlation analysis provide reliable theoretical and data support for modeling and analyzing the vegetation index and carotenoid content.Regression analysis and neural network inversion modeling analysis of vegetation index and original spectral principal components with carotenoid content using hyperspectral information,the accuracy of neural network inversion modeling is obviously better than that of regression analysis,and genetic neural network has the best inversion effect,the R~2 value of the training decision coefficient of the genetic neural network was more than 0.886,and the R~2 value of the test decision coefficient was more than0.874,which has reached a very strong correlation level.The inversion results showed that the carotenoid content of vegetation leaves was affected by many factors,the neural network modeling method with better nonlinear mapping ability can effectively improve the inversion accuracy.In this study,based on the original spectral data of carotenoid content to inverse modeling.It was founded that the accuracy of inversion can be improved by broadening the research scope of the spectrum and improving the inversion modeling method.The results of this study for plant leaves carotenoid content of fast,accurate and nondestructive detection to provide some support of theoretical and technical.