Abstract：To investigate the accuracy of various spectral data preprocessing and feature information extraction methods,two kinds of widely grown strawberry of ″Toyonoka″ and ″Jingyao″ in Hubei province were taken as research objects to determine sugar value and moisture content. The strawberry hyperspectral data was collected by hyperspectral imaging system, and the reflectivity of original spectrum was corrected. Various spectral data preprocessing and feature information extraction methods were adopted to analyze the strawberry hyperspectral data and obtain optimum method. The results show that the best spectral preprocessing method for Toyonoka strawberry moisture spectrum is the combination of Savizky Golay smooth, 1 derivative technique and standard normal transformation. The best spectral preprocessing method for Toyonoka strawberry sugar spectrum and Jingyao strawberry moisture spectrum is the combination of moving average smoothing, 2 derivative technique and multiple scattering correction. The combination of spectral difference analysis and correlation coefficient method can realize the accuracy of the correlation coefficient method and the intuitive nature of spectral difference analysis, and can avoid complex calculations. The representativeness of selected wavelengths is ensured,and the spectral dimension is reduced from 520 to around 80.
2017, Vol. 38 
