Utilization of a portable Raman spectrometer combined with a PCA-SVM model for starch type differentiation

Abstract

This article explores the feasibility of non-destructive rapid detection of various starch types using Raman spectroscopy and machine learning models. Raman spectra of 89 samples of corn starch, 88 samples of wheat starch, and 86 samples of cassava starch were obtained using a self-developed portable Raman spectrometer, Hx-Spec. Rapid discrimination models were established using Principal Component Analysis (PCA) and Support Vector Machine (SVM) to identify the categories of corn starch, wheat starch, and cassava starch. After optimization, the accuracy of the training set was 91.84%, and the accuracy of the test set was 93.67%. The study demonstrates that the self-developed portable Raman spectrometer, in conjunction with the PCA-SVM analysis method, can rapidly and non-destructively obtain starch molecular fingerprint information and accurately discriminate category attributes.