Wavelet Denoising Method Research of Soybean Straw Cellulose Near Infrared Rapid Detection

Abstract

In this paper,we made a research for soybean straw hemicellulose rapid detection by establishing a quantitative analysis model based on near-infrared spectroscopy. At first,146 samples were collected from varieties of soybean straws are gathered in different areas of Heilongjiang province, then made chemical testing of components and spectral scanning to soybean straw, the 140 samples were classified to two groups, in which 100 samples were chosen as calibration set and the remaining 40 samples were chosen as verification set. Wavelet transform was used to deal with the noise spectrum, selected DBN wavelet, Haar wavelet and Symlet wavelet in different layers under penalty threshold, Bridge-massart threshold, and default global threshold for spectral signal decomposition and reconstruction, compared with other traditional noise reduction methods,Symlet2-2 layer decomposition wavelet basis for hemicellulose spectral processing possessed better effect with the determination coefficient of validation set rising from 0.462524 to 0.6314158 after processing.