Identification of zinc pollution in rice plants based on two characteristic variables.

Traditional chemical methods used to measure the zinc content in rice plants are time-consuming, laborious, requires reagents, and have a limited monitoring range, while the Raman spectroscopy method has the advantage of being fast, non-destructive, and requires no reagents. Unfortunately, the identification accuracy of the Raman partial least squares (PLS) model based on principal components is only 53.33%. To boost this, a One-Way ANOVA method was used to extract the characteristic variables in the Raman spectra. Based on these Raman variables, a model for identifying zinc stressed samples was established. The identification accuracy was improved to 70% but still fell short of the measurement requirements. To further enhance these results, the Raman spectrum was decomposed into components based on the Hilbert Vibration Decomposition (HVD) method. Using characteristic variables of the Raman spectrum and its HVD components to establish a PLS model, the identification accuracy of the test set is raised to 90.25%. These results are a significant improvement from those obtained using a model solely based on the Raman spectral characteristic variables, revealing that HVD components provide highly effective identification information. A Raman modeling method based on the characteristic variables of the HVD component is an innovative way for improving the accuracy of Raman detection, especially for the measurement of trace substances.