Leaf Area Index (LAI) is an important indicator of the growth of crops. Timely and accurately obtaining the tobacco leaf area index has important research significance for assessing the growth of tobacco and optimizing the field management system. This article takes Tobacco in Xuanzhou District, Anhui Province as the research object. Based on the Sentinel-1 radar data and Sentinel-2 multi-spectrum data, combined with ground field survey data, based on statistical models and Prosail models, Tobacco Wang is long-term and mature LAI. Compare the advantages and disadvantages of 2 counter -trumpet methods. The following conclusions are mainly drawn:
(1) Based on the method of leaf area indexes based on the statistical m-odel, through the OOB(Out-OF-Bag Data,OOB),17vegetation indexes with th-e importance score more than 50 points are preliminarily selected, and at this basis Using SPA(Susscesive projections algorithm,SPA) to further select features. Based on the principle of minimum the principle of the minimum equity, the top 10 vegetation indices and Sentinel-1 radar data VV, VH polarization backscatter coefficient data were selected as independent variables to construct the tobacco leaf area index inversion model. Veority Index-VV combination has the highest counterpart accuracy, and the countermeasure accuracy of vegetation index, vegetation index -VH, vegetation index-VH, and VH gradually decreases.
(2) Among the leaf area index inversion algorithms based on statistical models, random forest (RF) is used to invert the highest accuracy, R2 up to 0.7992, RMSE(Root mean square error,RMSE) is 0.21; BP neural network (BP) is followed by R2 up to 0.7718, RMSE is 0.22; and finally radial basis function neural network (RBF), R2 up to 0.77, RMSE 0.25.
(3) Based on the leaf area index inversion method of PROSAIL model, Sentinel-2 data blue, green, red and near-infrared bands were selected to construct a lookup table, and on this basis, tobacco leaf area index inversion was carried out and accuracy analysis was carried out. The results showed that the R2 was up to 0.8083 and the RMSE was 0.316 when the tobacco leaf area index was inverted using the physical model.
谷歌
火狐
