Estimating Winter Wheat LAI Using Hyperspectral UAV Data and an Iterative Hybrid Method

Abstract

Leaf area index (LAI) is an important indicator 1 for crop growth monitoring. Due to the small number of 2 ground-measured samples, the hybrid method, using the radiative 3 transfer model (RTM) to generate simulated samples and com-4 bining with the regression model, is popular for LAI estimation. 5 However, there is still difference between simulated spectrum 6 and measured spectrum, which may affect the inversion results. 7 In this study, an iterative hybrid method combines BP neural 8 network, PROSAIL model and an optimal sample selection 9 method for crop LAI estimation was proposed. A small number 10 of ground-measured samples and unmanned aerial vehicle (UAV) 11 hyperspectral data were used to estimate LAI firstly. Then the 12 initial LAI result was used as the parameter of PROSAIL model 13 to generate the simulated spectrum. The simulated spectrum with 14 high similarity with the UAV spectrum and corresponding LAI 15 value would be treated as new samples for BP neural network. 16 After several iterations, a reasonable sample set was obtained to 17 estimate winter wheat LAI. The method proposed in this study 18 is evaluated using ground-measured test samples and compared 19 with the common hybrid methods. Results indicate that with 20 the increase of the number of training samples, the accuracy 21 of estimation model is improved (RMSE/MAE decreased from 22 0.4685/0.0301 to 0.4377/0.0272, respectively, while R 2 increased 23 from 0.5857 to 0.6384). Also, the accuracy of proposed iterative 24 hybrid model is higher than that of commonly used hybrid model. 25 The experiments demonstrate the relatively high accuracy of 26 the proposed iterative hybrid method, which could be used for 27 vegetation parameter estimation with only a small number of 28 ground samples.