Study On The Hyperspectral Estimation Model For Grain Crude Protein Content Of Indica Rice With Different Climatic Ecotypes

To find the optimal growth period and sensitive wavelengths of crude protein content in grain of indica rice with different climatic ecotypes,and the characteristic wavelengths and optimum model for spectroscopic determination of crude protein content in indica rice grains.Multiple varieties of early indica rice,middle-season indica rice and late indica rice in the middle and lower reaches of the Yangtze River in 2019 were selected as the research objects.The canopy spectral reflectance data of early,middle and late indica rice at key growth stages were collected by Hand Held 2 spectrometer,and the grain spectral reflectance data were collected by Field Spec 3 spectrometer after the rice were harvested.Through analyzing the correlation between spectral reflectance and the crude protein content in grains,the prediction models for the crude protein content of indica rice grains with different climatic ecotypes based on the canopy spectral reflectance and the estimation models for the crude protein content of indica rice grains based on the grains spectral reflectance were established,and the following results were obtained:1.The spectral reflectance of early indica,middle indica and late indica rice cultivars with different genotypes showed the same trend,but the reflectance intensity was different.And it’s specifically showed that the visible band difference is small,while the near infrared band difference is obvious.2.The correlation between the first derivative spectral reflectance of the canopy and the crude protein content of the grain at 514,580,638 and 695nm reached an extremely significant level at booting stage in the early indica rice researches;among the estimation models based on sensitive wavelength,the quaternary linear model had the best estimation effects,with the calibration set up R² and RMSE were 0.566,0.342%,and the validation set up R² and RMSE were 0.518,0.154%,respectively;in the estimation model that constructed based on the spectra parameters,the polynomial model which constructed by NDSI(R′₉₄₁,R′₉₀₇)as the independent variable had a better estimation effects,and its calibration set R² was 0.728,RMSE were 0.270%,respectively,the validation set R²,RMSE and RPD were 0.664,0.476%and 0.821,respectively.3.The correlation between the first derivative spectral reflectance of the canopy and the crude protein content of the grain at 567,585,633 and 722nm reached an extremely significant level at waxen ripen stage in the middle-season indica rice researches;among the estimation models based on sensitive wavelength,the three-dimensional linear model had the best estimation effects,with the calibration set up R² and RMSE were 0.613,0.430%,and the validation set up R² and RMSE were 0.634,0.296%,respectively;in the estimation model that constructed based on the spectra parameters,the polynomial model which constructed by DSI(R′₆₀₃,R′₅₈₂)as the independent variable had a better estimation effects,and its calibration set R² was 0.741,RMSE were 0.352%,respectively,the validation set R²,RMSE and RPD were0.640,0.304%and 1.600,respectively.4.The correlation between the first derivative spectral reflectance of the canopy and the crude protein content of the grain at 514,580,638 and 695nm reached an extremely significant level at starting ear stage in the late indica rice researches;among the estimation models based on sensitive wavelength,the quaternary linear model had the best estimation effects,with the calibration set up R² and RMSE were 0.549,0.320%,and the validation set up R² and RMSE were 0.547,0.311%,respectively;in the estimation model that constructed based on the spectra parameters,the polynomial model which constructed by RSI(R′₈₃₇,R′₄₄₈)as the independent variable had a better estimation effects,and its calibration set R² was 0.653,RMSE were 0.281%,respectively,the validation set R²,RMSE and RPD were 0.559,0.308%and 1.326,respectively.5.The spectral reflectance decreased with the increase of crude protein content,roughly showing the rule of medium indica>late indica>early indica;in the DSI,NDSI and RSI optimal spectral index models that constructed with any two wavelengths of original and first derivative reflectance,the PLSR model had a good effect,the calibration set R² and RMSE are0.841 and 0.507%,and the validation set R²,RMSE and RPD are 0.81,0.542%and 3.41,respectively;among the full-wavelength model,the effect was shown as PLSR>SMLR>PCR,the PLSR model established based on the original spectrum had the best effect,with the calibration set R² and RMSE are 0.867 and 0.464%,and the validation set R²,RMSE and RPD are 0.856,0.472%and 3.94,respectively;during the characteristic wavelength models,the model constructed by the first-order differential is better than the original spectrum,especially the PLSR model based on the first derivative spectral reflectance is more stable,the training set R² and RMSE are 0.842 and 0.506%,and the testing set R²,RMSE and RPD are 0.823,0.523%and 3.39,respectively.