Research On Modeling And Inversion Of Winter Wheat Leaf Nitrogen Content Based On UAV Hyperspectral Remote Sensing

Since the end of the 20th century,the unmanned aerial vehicle(UAV)remote sensing was rapidly rised and flourished,and it had becomed one of the leading aviation remote sensing technologies in current and future.So,the UAV remote sensing had broad prospect of application.In recent years,as the imaging hyperspectral remote sensing technology rapid development and wide application,the imaging hyperspectral remote sensing based on UAV had characteristics such as high accuracy,the imaging continuously and quickly.So,it had caused a research upsurge both in China and abroad.In this paper,the leaf nitrogen content of winter wheat was taken as the research object,and the ground plot test of winter wheat was carried out in 2014-2015.Near-surface winter wheat hyperspectral data was acquired using a drone-mounted imaging spectrometer and a ground-based non-imaging spectrometer.Combining with ground measurement data,the UAV spectroscopy characteristics of winter wheat were analyzed to extract the characteristic spectrum and sensitive spectral parameters of the response leaf nitrogen content of UAV in winter wheat.Comprehensively using multivariate regression,random forest,BP neural network and other methods,an estimation model for the leaf nitrogen content of winter wheat was established,and the optimal model was used to perform remote sensing inversion of the UAV hyperspectral imagery.On this basis,the relationship between leaf nitrogen content and nitrogen status of winter wheat was analyzed.An inversion model of nitrogen nutrient index of winter wheat was established and remote sensing image inversion was carried out to achieve a rapid diagnosis of nitrogen status of winter wheat.The paper mainly developed in following several aspects works and had obtained the innovation research results and the contribution:(1)Verified the effectiveness and availability of high spectral data obtained by UAVThe validity of hyperspectral data acquired by UHD185 sensor on UAV was verified.Based on the ground synchronous measured hyperspectral data,the accuracy and validity of UAV hyperspectral data was tested by using the correlation coefficient and spectral index calculation method.And based on the simulated reflectance spectrum data of PROSAIL vegetation radiation transmission model,two aspects of different growth period and different nitrogen application level were tested again.The results show that the UAV hyperspectral data had high data sampling accuracy and high reliability.(2)Analysis of UAV hyperspectral characteristics of winter wheat From different aspectsAccording to the UAV hyperspectral data,characteristics of canopy spectra of winter wheat were analyzed from four aspects:different growth stages,different nitrogen levels,different water conditions and different leaf nitrogen contents.The spectral characteristics of winter wheat at different growth stages were different,but the overall trend was the same.In the visible region,the canopy reflectance of winter wheat decreased slightly from jointing stage to flag leaf stage.From flag leaf stage to filling stage,the canopy reflectance of winter wheat increased continuously,and the reflection peak at 550nm became less,and tended to disappear.In the near infrared region,from jointing stage to flag leaf stage of winter wheat,canopy reflectance spectra showed an increasing trend;from flag leaf stage to flowering stage,canopy reflectance decreased gradually and tended to be stable;from the flowering stage to the filling stage,the spectral red edge position shifted to short wavelength,and the phenomenon of "blue shift"appeared.The UAV hyperspectral differences of winter wheat under different nitrogen levels were obvious.The near infrared region difference was more obvious than the visible region.The near infrared band had more sensitivity to nitrogen fertilizer application.With the increase of nitrogen application rate,the spectral red edge position shifted to long wavelength,and the phenomenon of "red shift" appeared.Under different water conditions,the difference of hyperspectral curves in the near infrared region of winter wheat canopy was greater than that in the visible region.In the visible range,with adequate irrigation water,the reflectance of Winter Wheat Canopy Spectra decreased,while in the near infrared region,winter wheat canopy spectral reflectance showed a rising phenomenon and the difference was significant.The results showed that the near infrared spectrum of canopy biomass of winter wheat was more obvious with the increase of soil water conditions.The spectral characteristics of Winter Wheat with different nitrogen content were different.In the visible region,the reflectance of the spectrum was negatively correlated with leaf nitrogen content.The lower the leaf nitrogen content was,the higher the spectral reflectance.In the near infrared region,the reflectance of the spectrum was positively correlated with leaf nitrogen content,and the reflectance increased with the increase of the leaf nitrogen content.(3)Proposed a hyperspectral representative band extraction method based on band correlation threshold in response to winter wheat leaf nitrogen contentAccording to the UAV hyperspectral data,the first order differential and continuous removal methods were used to extract the representative bands of LNC,and the results of the extraction band were modeled,analyzed and verified.A hyperspectral dimension reduction method was proposed based on correlation threshold analysis.Using this method,the representative bands combination of four key growth periods of LNC was obtained.Analyzed and compared the three results,which based on the hyperspectral correlation threshold was better,the number of bands was much less and the band was more representative.From the results of modeling,the R2 values of jointing stage,flag leaf stage and flowering stage were greatly improved.(4)New spectral indexes were constructed based on the band correlation threshold,the estimation of LNC of winter wheat at different growth stages of UAV hyperspectral period was achieved.The present spectral index was improved and optimized based on the representative band extracted by the band correlation threshold method in response to LNC of different growth stages of winter wheat.The ratio spectrum index,the difference spectral index and the normalized spectral index were constructed by the two combinations of the representative bands.On this basis,the 450nm band was introduced to modify the newly constructed spectral index.A new model for the LNC estimate in different growth stages was established based on the newly constructed spectral indexes.The spectral indexes constructed from different band combinations and sensitive bands were obtained in response to LNC in different growth stages.Compared with the existing modeling results of nitrogen spectrum index,the precision of the model had been improved,and the improvement of the flag leaf stage was the most obvious.The introduction of the 450nm correction band was of great help to the improvement of the modeling precision by the newly constructed spectral index during the flowering and filling stage.(5)Monitoring of LNC and preliminary diagnosis of nitrogen nutrition status of Winter Wheat based on UAV hyperspectral imagingBased on the representative bands extracted by the method of band correlation threshold,the hyperspectral quantitative model of UAV was established by multiple linear regression and random forest and BP neural network responding to LNC.The modeling results were compared and analyzed.Remote sensing image inversion is realized by selecting the optimal result model for remote sensing mapping.The change of nitrogen nutrition index of Winter Wheat under different nitrogen level was analyzed,and the estimation model of LNC and nitrogen nutrition index of winter wheat was established.The model was analyzed and verified,and remote sensing image inversion was carried out to realize the remote sensing monitoring of nitrogen nutrition status of winter wheat.This research has successfully carried out the data processing and application of UAV imaging hyperspectral data,and provided a method of UAV remote sensing image quantitative remote sensing application,which has important theoretical and practical significance.