Research On The Method Of Remote Sensing Monitoring Of Leaf Area Index Of Rice And Wheat In The Major Fertility Stage

Rice and wheat are major global food crops that require accurate monitoring and management due to the complexity of their growth characteristics.Effective monitoring of the growth status of rice and wheat facilitates the detection and treatment of abnormal plant growth and the timely adoption of appropriate management measures,such as fertilization,irrigation,pest and disease control,to ensure the growth and development of rice and wheat.LAI is the ratio of crop leaf area to crop planting area,which is an important index to measure the growth status and yield of crops.LAI obtained by remote sensing technology can reflect the crop growth status,including information on growth cycle,leaf coverage degree,growth rate and biomass,which provides a quantitative and visual guidance basis for agricultural production management.This study explored the spectral characteristics and leaf LAI of rice and wheat during different months throughout the entire growth period,starting from two scales of geospatial hyperspectral imaging technology and satellite remote sensing technology.Image processing techniques,spectral analysis techniques,and mathematical statistical methods were used to respectively study the correlation between LAI and leaf hyperspectral reflectance,crown multispectral reflectance,and surface reflectance of rice and wheat.The research results show that there is a certain correlation between the LAI of rice and wheat and leaf hyperspectral reflectance,crown multispectral reflectance,and surface reflectance.By establishing a quantitative analysis model,accurate estimation of LAI can be achieved.The details of the study and the results of the study are as follows:(1)In the geospatial hyperspectral scale study,the reflectance spectra of rice crowns during the elongation stage,booting stage,heading stage,and post-flowering stage were first obtained using an ASD spectrometer,and the correlation between the original reflectance(OR),first-order derivative transformation(FD),reciprocal transformation(l/R),logarithmic transformation(LOG),and LAI was analyzed.Then,feature bands of the spectral data were selected based on the successive projection algorithm(SPA)and Pearson correlation method.In addition,estimation models based on feature bands and vegetation indices were established using ridge regression(RR),partial least squares(PLS),and multiple stepwise regression(MSR).The results show that the correlation between crown reflectance spectra and LAI was significantly improved after FD transformation.The modeling effect of using SPA to select FD feature bands is better than that of using Pearson correlation.The best modeling combination is FD-SPA-VI-RR,with a determination coefficient(R2)of 0.807 and root mean square error(RMSE)of 0.794 for the training set,R2 of 0.878 and RMSE of 0.773 for validation set 1,and R2 of 0.705 and RMSE of 1.026 for validation set 2.The results show that this model can accurately predict the growth of rice,meeting the requirements of large-scale statistical monitoring of rice growth indicators in the field.(2)In the satellite multispectral scale study,based on the leaf LAI of wheat during the elongation stage,booting stage,and flowering stage as the research basis,corresponding satellite images were downloaded,vegetation indices were calculated,and the correlation between spectral parameters and LAI was analyzed.Based on this,BP neural network,random forest(RF),and support vector machine(SVM)models were established to estimate LAI.The genetic algorithm(GA)optimization algorithm was further introduced to combine with the three models and compare whether GA improves the model accuracy.The results showed that the GA-RF model performed best in three fertility periods of wheat,with R2 of 0.863 and RMSE of 0.191 for the modeling set at the nodulation stage,and R2 of 0.683 and RMSE of 0.302 for the validation set.R2 of 0.785 and RMSE of 0.269 for the modeling set at the gestation stage,and R2 of 0.657 and RMSE of 0.339 for the validation set.The modeling set for the flowering stage had R2 of 0.790 and RMSE of 0.231;the validation set had R2 of 0.656 and RMSE of 0.322.This study provides quantitative and accurate monitoring of rice and wheat LAI at two different remote sensing scales,which provides a theoretical basis for the study of rice and wheat longevity.Therefore,the star-land remote sensing monitoring method proposed in this study can provide an important reference basis for agricultural production,which is of great significance for improving agricultural production,increasing grain yield and ensuring food security.