Monitoring Of Winter Wheat Growth Indicators Based On Different Spectrometers

Winter wheat is an important grain crop in China,and its yield and quality are closely related to national food security and the formulation of relevant regulations.Therefore,real-time and accurate monitoring of winter wheat growth over a large area is of great significance for field precision management and sustainable agricultural development.Remote sensing technology has been widely used in the agricultural field because of its advantages of rapid and non-destructive large-area monitoring.In this study,winter wheat with different ecological sites,different nitrogen treatments and different varieties in Henan Province was selected as the research object.Spectral data and agricultural parameters of winter wheat CGMD portable crop growth monitoring and diagnosis instrument and ASD high spectrometer were obtained respectively,and the growth indicators(LNC,LAI and LDW)of winter wheat were monitored based on the two spectrometers.Firstly,the reliability of CGMD data was verified,and the localization model and comprehensive model of winter wheat growth index at different ecological points were constructed by vegetation index.Secondly,based on ASD hyperspectral data,using different spectral preprocessing combined with feature variable extraction,combined with vegetation index and a variety of machine learning algorithms,a hyperspectral monitoring model of winter wheat growth indicators under the influence of multiple factors was constructed.The results can provide a basis for the monitoring of winter wheat growth indicators and field fertilization management in Henan Province.The main conclusions are as follows:(1)The variation characteristics of winter wheat agronomic parameters in four different ecological points of Huaxian,Xuchang,Fangcheng and Yuanyang were studied.The results show that the variation characteristics of the three types of agronomic parameters have obvious regional differences.On the one hand,there are differences in the content range,and the variation intervals of agronomic parameters in different ecological points are different.The variation range of winter wheat LNC was 1.52%-4.55%,1.42%-4.15%,1.06%-4.29% and 0.24%-3.67% in four different ecological sites in Huaxian,Xuchang,Fangcheng and Yuanyang,respectively.The content ranges of winter wheat LAI in four ecological sites were 1.37-9.63,0.95-9.35,0.87-9.05 and 0.16-8.58,respectively.The content ranges of winter wheat LDW in four ecological sites were 63.88 g/m2-437.12 g/m2、47.07 g/m2-431.69 g/m2、41.36g/m2-429.69 g/m2 and 10.68 g/m2-422.19 g/m2,respectively.According to the size of the endpoint,it can be considered that Huaxian > Xuchang > Fangcheng > Yuanyang.On the other hand,due to the difference in the change trend,with the increase of nitrogen application rate,the change trend of agronomic parameter content of each ecological point was basically the same,which was positively correlated with the nitrogen application rate.With the advancement of the growth period,the agronomic parameters of Yuanyang winter wheat gradually decreased,while the agronomic parameters of other ecological points showed a trend of "low-high-low".(2)A monitoring model of winter wheat growth indicators based on CGMD portable crop growth diagnostic monitor was established.Firstly,the reliability of CGMD data is verified by using the vegetation index constructed by ASD spectrometer corresponding to the band.The results show that the determination coefficient of vegetation index constructed by the two spectrometers is above 0.7.The results showed that the vegetation index obtained by the two spectrometers was highly correlated,and the CGMD spectrometer could be used to monitor the growth index of winter wheat.Secondly,the vegetation index was used to construct a localized model and a comprehensive model for monitoring winter wheat growth indicators.The results showed that the vegetation index model constructed based on CGMD data had certain monitoring ability for winter wheat growth indicators.The accuracy of the model built by different ecological points is different,and the accuracy of the comprehensive monitoring model built by multiple ecological points is not as good as that of the model built by single ecological point.All the winter wheat LNC estimation models were better performed by NDVI model.The validation set Rv2 of the comprehensive model constructed based on NDVI was 0.62,and the Rv2 of the single ecological point NDVI model was 11.3%-24.2% higher than that of the comprehensive model.The established LAI models of winter wheat all performed well with RVI model.The validation set Rv2 of the comprehensive model constructed based on RVI was 0.61,and the Rv2 of the single ecological point RVI model was1.6%-8.2% higher than that of the comprehensive model.The winter wheat LDW estimation models were all better performed by NDVI model.The validation set Rv2 of the comprehensive model constructed based on NDVI was 0.61,and the single ecological point NDVI model Rv2 was 3.1%-10.9% higher than that of the comprehensive model.(3)Combined with various methods,a monitoring model of winter wheat growth indicators based on ASD hyperspectrometer was established.The original spectra were pretreated by LOG,continuum removal CR and standard normal SNV,and the characteristic variables of agronomic parameters were extracted by vegetation index VI,competitive adaptive reweighting algorithm CARS and discrete wavelet transform DWT.The monitoring model of winter wheat growth indicators was constructed by vegetation index VI,support vector machine SVM,K-nearest neighbor KNN and Bagging algorithm.The results showed that compared with the original canopy spectra,the correlation between the pretreatment spectra and LNC,LAI and LDW of winter wheat was improved.The SVM model established by the wavelet approximation coefficient extracted by discrete wavelet transform by continuous removal spectrum was the optimal estimation model for winter wheat LNC.The modeling set RC2 was 0.84 and the verification set RV2 was 0.83.The SVM model constructed based on logarithmic spectroscopy and the feature bands extracted by competitive adaptive realgorithm is the optimal estimation model of winter wheat LAI.The modeling set RC2 was 0.81,and the verification set RV2 was 0.81.The Bagging model constructed based on the discrete wavelet approximation coefficient of continuum removal spectrum is the optimal estimation model of winter wheat LDW.The modeling set RC2 was 0.91,and the verification set RV2 was 0.80.