Study On Physiological And Biochemical Parameters Of Rice Based On Hyperspectral Remote Sensing In Yellow Rive Irrigation Area Of Ningxia

Hyperspectral remote sensing can provide the data that the precision agriculture required timely and accurately,it is one of the most fundamental technology of the development of the precision agriculture.It can be achieved that monitor the growth of vegetation mathematical model that was established by studying the relationship between hyperspectral data and physiological and biochemical parameters of vegetation.In this study,we obtained the physiological and biochemical parameters(chlorophyll content and leaf area index),ground hyperspectral data and remote sensing images of low altitude simultaneously in Ningxia Yellow River Irrigation Area.Firstly,we studied the spectral characteristics of rice under different growth stage and different fertilization levels,and analyzed the relationship between the spectral reflectance of rice canopy and the parameters.Then,we constructed the estimation models of SPAD and leaf area index(LAI)which were based on characteristic bands and BP neural network(BPNN)by extracting the best characteristic bands and the band combinations of spectral index of rice canopy spectrum.Finally,we generated the distribution map of SPAD and LAI of rice by combining the estimation models and the UAV hyperspectral images,the estimation ability of different models to them were compared,respectively.The main results of this study are as follows:(1)The spectral characteristics and the characteristics of red edge at canopy and leaf under the different growth stages of rice and different fertilization levels were analyzed.The result showed that the spectral reflectance decreased gradually with the increase of chlorophyll content under the same carbon level in the visible bands,and the opposite regularity in the near infrared band.At the same nitrogen level,with the increase of carbon content,the spectral reflectance hasn’t show changes obviously in the visible range but significantly changed in the NIR range,however,it would decreases when the carbon reaches a certain level.The spectral characteristics at different growth stage of rice were extremely different,the spectral reflectance of rice canopy reached the lowest at heading stage and the highest at milky stage in visible band which was exactly the opposite in the NIR range.The red edge parameters of rice showed a certain regularity with the growth period,from jointing stage to heading stage,the position of red edge shifted to longer wavelength but to shorter when it was at milky stage.The area and amplitude of red edge increased firstly and then decreased with the growth of rice,the maximum value of whom were obtained at the heading stage.(2)The results suggested that the SPAD value of rice was the largest at the jointing stage among the whole growth stage after studying the changes of chlorophyll content at the development of rice.It was found that the correlation between the canopy spectral reflectance and SPAD of rice at its different growth stages was regular.The correlation coefficient in the visible bands increased gradually with the growth of rice,while the correlation decreased.In the NIR bands,the correlation decreased along with the correlation coefficient.Characteristic bands and 6 kinds of spectral index(SI)which composed by different bands at each growth stage of rice were choose to construct the estimation model of rice SPAD,the results show that the model which based on spectral index(SI)and BP neural network was superior to the model that based on characteristic bands.(3)The leaf area index(LAI)showed a trend of increasing firstly then decreasing with the growth of rice,and reached the maximum value at its heading stage.The correlation between canopy spectral and LAI of rice in the visible bands is higher than it in the NIR bands at the same period,while it decreased both in the visible bands and NIR bands at different stage.Selected characteristic bands and SIs which formed by different bands at different growth stage of rice,and then constructed two kinds of models to establish the LAI,respectively.It shows that the model based on spectral index and BP neural network can improve the accuracy of estimation respectively.(4)With the estimation models based on characteristic bands and BP neural network of SPAD and LAI and UAV hyperspectral images,simulated the SPAD and LAI at jointing stage of rice and then the distribution maps of SPAD and LAI obtained,respectively.The results show that the distribution of LAI and SPAD which estimated by SIs and BP neural network fit the actual situation in good agreement with higher feasibility.