Hyperspectral Characteristics In Main Growing Stages And Estimating Models About Grain Protein Content Of Winter Wheat

With the improvement of standard living, people’s demand of wheat has changed from quantity to quality. Producers, government departments and relevant experts have attach more and more attention to predicting grain protein content in winter wheat. Wheat protein contents has high correlation with nutritional quality and processing quality. So grain protein content is often be seen as one of the main factors that affect wheat quality. Remote sensing is a fast, non-destructive, real-time and dynamic way to monitoring crop physiological and biochemical characteristics of large area research and it has been widely applied in monitoring the grain protein content of wheat in recent years. With the development of hardware technology level, one important research area of using remote sensing technology to monitor quality is to improve the stability of the interannual model, explore the optimal monitoring index and the optimal methods of modeling.In order to establish grain protein content inversion model which is universal and easily be extensionon, The study based on the data of three years at Beijing Precision Agriculture Demonstration Base of Xiaotangshan,2009-2010datas are used to analysing and modeling,2008-2009and2011-2012datas are used to validate the model.1) Chosen wheat canopy spectral data under different growth periods, different varieties, different dates of seeding and different nitrogen treatments, picked up the main characteristics of the spectral, analyzed the spectral characteristic curve under effects of various treatments, and studied the relationship between the spectral curve features in those treatments and GPC in maturation stage. The results shows that:from reviving stage to maturation stage, in the350-750nm band, the value of wheat canopy reflectance is the lowest in heading stage. In the750-1300nm band, it’s the highest reflectance. In the late growth periods of winter wheat, especially during the filling stage, the canopy spectral characteristics is significantly correlated with grain protein content, filling stage is the optimal period to monitoring GPC; The disparities caused by varieties and sowing are the smallest at optical wavelengths in anthesis;. The difference caused by nitrogen fertilization is widening gradually after fertilize, and formed obvious gradient; The correlation between GPC and spectral characteristic curve has some different under different varieties and different sowing dates. The study analyzed the spectral characteristic curve under different treatments and its relationship with GPC. To provided the reference for selecting suitable vegetation index to monitoring.wheat quality.2) Using single spectral index to monitoring wheat grain protein is a common method in remote sensing, while considering the relationships nitrogen, water, growing with grain quality of integrated spectral indices is effective to improve the precision of monitoring model. Firstly, select spectral indexes by the data of2008-2009which were bound up with leaf nitrogen content, water content of leaf and leaf area index respectively. Analysis the relationships between single spectral index and integrated spectral indexes with GPC respectively. Select the integrated model of high accuracy and test it by the data of2009-2010. To determine the best annual performance stable model and using the temperature factor to adjust it. The results showed that: comprehensive indexes model to monitoring GPC is significantly higher precision than that of the single one, and between annual reached the significant level.The model of R2is0.831, RMSE is1.557%, show that there is higher degrees between the predicted values and the observed value. The result can provide the theory basis and the technical support to monitor grain protein content by remote sensing.3) More works were needed to increase the accuracy and interpretability of prediction model of wheat grain protein content. The research based on the relationships between GPC nitrogen-chlorophyll, selecting proper vegetation index by chlorophyll and using partial least squares regression (PLS) method to build monitor model. The results show that flowering period is the optimal period for GPC monitoring. At flowering stage, there is a high correlation between nitrogen and chlorophyll with the corresponding density. Five indexes are chosen to build the model to calculating GPC by chlorophyll.. Coefficient of determination is0.77, RMSE is0.95%. For validating the model by the data of other year, RMSE is1.22%. The model shows high precision and stability in different years.4) In order to improve the stability of the model, every vegetation index is been adjusted by the mean of modeling vegetation index and validating vegetation index before using. This method provides a reference for increasing modle expansion of time and regional. Strengthen the consistency of spectral indices in different years, and then enhance the spatial and temporal stability of the model.