Inversion Method Of Chlorophyll Content In Rice Leaves Based On Hyperspectral

The high-spectral monitoring technology can rapidly and accurately supply the data information used for precision agriculture and promote the development of the precision agriculture.By using the correlation between the high-spectral monitoring data and the plant biochemical parameters,the mathematical function model of the relation between the two plants is constructed,and the inversion and the monitoring of the plant parameters are completed.And provides a certain theoretical basis for the future high-spectrum monitoring of the chlorophyll content of the rice leaves.In this paper,the high spectrum of rice leaves and the content of chlorophyll in rice leaves were obtained by obtaining the high spectrum of rice leaves by taking the high spectrum of rice leaves.In this paper,the spectral changes of the leaves of rice in the stage,the jointing stage and the heading stage of rice were analyzed.In this paper,the characteristics of the variation of the leaf spectrum under different nitrogen levels under different nitrogen levels are selected,and the best characteristic band combination,the best vegetation index and the optimal"trilateral"parameter of the high spectrum are selected as the input feature of the subsequent inversion of the chlorophyll content model.Based on the three methods of BP neural network(PLSR),BP neural network(BP)and PSO(PSO),the model of chlorophyll content inversion is constructed,and the effective model verification method is used to evaluate the quality of the inversion model with the root mean square error(RMSE).And provides a certain theoretical support for the technical monitoring technology of rice health.The main contents of this paper are as follows:(1)In this paper,the original spectrum,red-edge spectrum and red-edge parameter of rice leaves at three growth stages(tillering stage,jointing stage and heading stage),different chlorophyll contents and different nitrogen levels were studied and analyzed.The results showed that the original spectrum and red-edge parameters of rice at different nitrogen levels were different in three growth stages(tillering,jointing and heading).The results showed that from tillering stage to heading stage,the spectrum of rice leaves in the near infrared band increased gradually,and maintained a trend of a platform.In the jointing stage,the amplitude and area of the red edge reached the maximum,and the red edge parameters showed obvious differentiation when the chlorophyll content was different.As the concentration of nitrogen fertilizer increased,the spectral reflectance of rice leaves decreased.The trend,and the position of the red edge also changed.So a series of changes of the spectrum of the rice leaves can play a key role in the monitoring of the growth and development of the rice.(2)Feature extraction for multi-dimensional high-spectrum data firstly,using SPA to select a high-spectrum sensitive wave band,a time division period,a jointing stage and a heading period spectrum,wherein the number of the sensitive wave bands is 3;then,the vegetation index is constructed by using the original spectrum of the leaves of the three growth period of the rice in a combination of any wave bands,And the correlation of the best vegetation index with the corresponding leaf chlorophyll value is more than 0.5,and the correlation of the vegetation index RVI with the chlorophyll content is the highest.The degree of correlation of the SDb,Sy,and?r 3 parameters in the"trilateral"parameters is the highest.(3)The obtained optimal high-spectral characteristic band,vegetation index and"trilateral parameter"are used as the input quantity of the chlorophyll inversion model of the BP neural network with the least square method,the BP neural network and the PSO.The research shows that the inversion effect is good in the three growing periods,the characteristic vegetation index in the high-spectrum characteristic is the best in the inversion of the chlorophyll content,the optimal inversion modeling method is the PSO-optimized BP neural network,the verification set R~2is 0.789,and the RMSE is 1.7689.It is possible to obtain the water quickly and accurately.Chlorophyll content in order to monitor the nutritional status of rice to provide some theoretical support.