Studies On The Optimum Choice Of Combination Wavelengths In Near Infrared Spectral Analysis For Sugarcane Clarified Juice

Using near-infrared spectral analysis techniques and chemometrics, included multiple linear regression (MLR),partial least squares (PLS) and artificial neural network (ANN),with the appropriate spectrum pretreatment methods, the quantitative analysis model of brix in sugarcane clarified juice was established. According to forecast results of Root Mean Square Error of Prediction (RMSEP), optimum spectra band and features wavelength were selected, it provided an important basis for the design of the small-scale near-infrared instrument for measuring brix in clarified juice. Firstly, The appropriate correction sets and training sets were determined after abnormal samples and similar samples were selected by used principal components -Markov method , then correlation coefficient and moving window PLS methods were used to choose optimum spectra band, and then the five wavelength points in the above spectra band was further selected ,which based on the forecast effect ,to create the quantitative models with MLR,PLS and ANN methods, the five wavelength points were selected at the way of discrete and interval. The results showed that principal components－Markov method was effectively selected abnormal samples and similar samples; correlation coefficient method was effectively selected spectra band; the effect of quantitative analysis model Established by five-wavelength combinations (discrete, interval methods can be ) was similar to the entire spectrum (400-2500 nm) model, it provided an important basis for the design of the small-scale near-infrared instrument for measuring brix in clarified juice.