Hyperspectral inversion of maize biomass coupled with plant height data

Abstract

AbstractBiomass is an important indicator in estimating the growth and yield of crops. Mastering the growth state of crops by evaluating and monitoring biomass is crucial. Correlation analysis was used to identify hyperspectral sensitive bands with the best correlation with biomass, and a hyperspectral vegetation index was constructed. Plant height data was coupled with hyperspectral reflectance and vegetation index data, and a biomass inversion model was constructed using partial least squares (PLS), support vector machine (SVM), and random forest (RF) regression analysis. The R2, root mean square error (RMSE), and normalized root mean square error (NRMSE) were calculated to verify the accuracy of the model. Plant height coupled with hyperspectral reflectance and vegetation index data, R2, RMSE, and NRMSE of the biomass model establishment and validation, compared with the hyperspectral data alone, improved model accuracy of the three algorithms for maize (Zea mays L.) biomass inversion coupled with plant height data with the relative improvement in accuracy of 13.58, 9.6, and 1.05%, respectively. Additionally, the model becomes more stable, overcoming the spectral saturation of canopy to a certain extent. It provides supporting information for accurate farmland planning, growth monitoring, and yield estimation.