Abstract: In this study, a rapid method using Fourier transform infrared and ultraviolet absorption spectroscopies coupled with data fusion was established for the identification of the geographical origin of Boletus tomentipes. The original spectra of 96 samples collected from different growing regions were preprocessed by multiplicative signal correction (MSC), standard normal variate (SNV) and second derivative (2D) to decrease the noise interference. The spectral information about the fingerprint characteristics was chosen for low-level data fusion, and the spectral information of variables important in projection greater than 1 was selected by partial least squares-discriminant analysis (PLS-DA) for mid-level data fusion. Then the single and fused spectral data were analyzed by PLS-DA and support vector machine (SVM). The prediction performance of PLS-DA and SVM was compared. The results showed that based on FTIR, UV, low-level data fusion and mid-level data fusion data matrixes, the prediction accuracies of PLS-DA were 56.25%, 56.25%, 62.50% and 81.25%, respectively, and the prediction accuracies of SVM were 90.63%, 65.63%, 87.50% and 96.88%, respectively, suggesting that mid-level data fusion was better than the other data sources and SVM was better than PLS-DA. In conclusion, SVM based on mid-level data fusion can be a rapid and effective method for the identification of the geographical origin of B. tomentipes that facilitates food quality monitoring.

Key words: data fusion, Boletus tomentipes, geographical identification, ultraviolet absorption spectroscopy, infrared spectroscopy