Volume Water Content Prediction of Variable Bulk Density Soil Based on Spectral Characteristics
In order to eliminate effects of soil bulk density variance and soil properties difference on soil water content forecast precision, a method was proposed to forecast soil volume water content based on near-infrared spectroscopy. This paper investigated relationship between volume water content and near infrared spectral reflectance characteristics in 900-2500nm band, used spectral parameters and support vector machine to built quantitative prediction model for three type variable bulk density soil volume water content, normalized signal characteristics by relative dry soil characteristic vector, putted forward three further processing methods. It was used support vector machine (SVM) method to establish spectral characteristics inverting soil volume water content model of undisturbed soil, and model parameters were optimized by genetic algorithm, through predict error comparison, the final determination was that relative characteristics variation of first-order derivative signal as model input characteristic vector, GA-SVM model prediction had best effect and its forecast error was 1.7866.