NIR Spectroscopy Based on DWT and LS-SVM for Prediction of Soil Moisture

Models of soil moisture prediction based on discrete wavelet transform (DWT) and least square support vector machines (LS-SVM) regression method were introduced in this paper. Applied Daubechies, Symlets and Coiflets wavelets at decomposing level of 4, the near-infrared spectra (NIRS) signals of 78 soil samples had been de-noised, and LS-SVM models were established and validated with 38 soil samples. It shows that db4 is the best. Within LS-SVM models established using db4 wavelet corresponding to different decomposing level, the best prediction effect were obtained when the decomposing level was 6 and wavelet was db4, which the correlation coefficient of the model is 0.9870 and the root mean square error for prediction (RMSEP) is 1.3628.