Research On Estimation Method Of Chlorophyll Content Of Northeast Japonica Rice Based On Low-altitude Remote Sensing Of UAV

Chlorophyll is one of the most important pigments used in light energy.The content of chlorophyll reflects the status of nitrogen nutrition,photosynthesis and heavy metal pollution in crops directly.Therefore,it is very important to estimate the chlorophyll content of crops accurately and rapidly.However the traditional chlorophyll measurement method is inefficient.Due to the satellite remote sensing and ground remote sensing established rice chlorophyll content inversion model.It's difficult to meet the needs of precision inversion of rice growth information at the regional level due to spatial resolution and spectral information limitations.However,with the development of UAV and hyperspectral technology,it provide the new method and technical support for solving the precise inversion of chlorophyll content in rice at the regional level.Below is the main research work:This study takes the northeast japonica rice as the main study object and carries out the rice field cultivation experiment based on the“3414”fertilizer design in the Daonan Experimental Field of Shenyang Agricultural University for two consecutive years from July 2015 to September 2016.Hyperspectral remote sensing imagery of rice canopy in the experimental field was obtained through a multi-rotor UAV hyperspectral remote sensing platform,and hyperspectral remote sensing classification methods Fisher Fisher,SVM,and second-generation wavelet decomposition algorithm were used to collect field hyperspectral data.The pure canopy hyperspectral image information was obtained from remote sensing images.The classification results showed that the average classification accuracy of the second-generation wavelet classification method was as high as 90.36%,which was significantly higher than other classification methods.Using the smoothing method and convolution smoothing method,the spectral curve is smoothed,and the normalization and detrending algorithm of variables are used for correction and transformation.The continuous projection method(SPA)and the random combination of two or two bands are used to construct the vegetation index.In the method of dimension reduction,the characteristic bands with large contributions to the chlorophyll content of rice were screened using the continuous projection method,and the two wavelength bands that did not meet the statistical basis were excluded.The final selected wavelengths were 606nm,648nm,539nm,677nm,465nm,712nm and 690nm.Three plant cover indexes were randomly constructed in the two bands of the whole band,namely the normalized vegetation index(NDVI),the ratio vegetation index(RVI)and the difference vegetation index(DVI),and finally screened out related to the chlorophyll content of rice.The best spectral bands for NDVI,RVI,and DVI were(512 nm,698 nm),(512 nm,697 nm)(525 nm,702 nm),and the correlations were 0.9033,0.9080,and 0.8021.The multivariate regression model,the BP neural model and the extreme learning machine(ELM)model were established by using the selected characteristic bands and vegetation indices and the chlorophyll content of rice,and the precision of the model was tested.The final determination of the optimal chlorophyll content inversion model is based on the extreme learning machine model of NDVI,DVI and RVI.The determination coefficient R~2 is 0.9174 and the simulation error RMSE is 0.7633,which is most suitable for the inversion of chlorophyll in Northeast Japonica rice.In this study,the hyperspectral information of the northeastern japonica rice canopy was obtained through a UAV hyperspectral remote sensing platform.An inversion model of hyperspectral information and rice chlorophyll content was established.The results can provide a theoretical basis and technical support for large-scale,rapid,non-destructive detection of chlorophyll content.