The Study On Hyperspectral Variables In Predicting Nitrogen Status,Pigments Content And Grain Protein Content In Rice

Crop nitrogen content is an important parameter in crop growth monitoring, crop yield estimation and quality assessment precision nitrogen management is of great importance in crop production. Traditionally, crop nitrogen nutrition diagnosis is based on chemical analysis of plant tissues, which are labor-intensive, time-consuming and involving hazardous chemicals. In recent decades, there are growing studies in assessing plant biophysical and biochemical parameters including plant nitrogen content by hyperspectral remotely sensed data. The objective of this study is to investigate the relationships among the plant biochemical parameters including plant nitrogen content, rice quality and plant hyperspectral reflectance in solution-cultured rice with different cultivars and nitrogen supplement levels in order to establish a hyperspectral approach in predicting rice quality by the hyperspectral reflectance.The spectral responses of the flag leaves to the nitrogen level were similar in the rice cultivars ("Xiushui110" and "Bing 9914") at different stages. The reflective value tended to increase with the stage over the full test wave band (350~2500nm). The spectral reflectance of the panicles were similar to that of the flag leaves in the visible light region, but they tend to decrease with the growth stages in the infrared wave band Some exiting derivatives including λ r, Dr, SDr, λ g, Rg and EGFN based on spectral position in the leaves of "Xiushui110" were found to change in consistent with the nitrogen levels at the different stages. The results showed that these spectral parameters could be used to distinguish the different nitrogen levels and the growth stages in rice.The original raw spectral data and the first derivatives spectral data were analyzed in correlation with the nitrogen content, the chlorophyll content and the carotenoid content in the flag leaves and in the grain husk in "Xushui110" at different stages to find out the wave bands sensitive to the change of these biochemical parameters and establish the estimation model between the spectral data and the biochemical parameters. The chlorophyll content is highly correlated with the nitrogen content in the rice crop, indicating that the nitrogen content can be predicted by the chlorophyll content indirectly. The regression models established in this study had been used to successfully predicted chlorophyl content and the carotenoid content in the flag leaves and the grain husks with promising accuracy in the solution-cultured rice.Protein and Starch are the dominant components in the rice grain, which determine the nutrition quality of rice. To predict the rice quality nondestructively, in time and at large-scale by remote sensing technique should be of great value. The correlations were analyzed between the hyperspectral data and the protein content as well as the amylose content in the rice grain and the results showed that the hyperspectral variables were strong related to the protein content in the grain while poor relating to the amylose content in the grain. The estimation model for the protein content in the grain for "XiushuillO" were: Pr%=4.902* e37l74*GNDV'+l 12.79 and Pr%=45.73*e24826*SRo+l 12.79. Using the estimating model to access the grain protein content in "Bing9914", there was high correlation between the estimated protein content and the measured protein content, with R2=0.99I4 and R"=0.9942 (n=4), respectively. The results suggested that the hyperspectral variables could be used to estimate the leaf chlorophyll content, the carotenoid content, the leaf total nitrogen content and the grain protein content in the solution-cultured rice.