A spectral characteristic analysis method for distinguishing heavy metal pollution in crops: VMD-PCA-SVM.

Abstract Exploring the characteristics and types of heavy metal pollution in crops has important implications for food security and human health. In this study, a method for distinguishing heavy metal-polluted elements in corn leaves was proposed. Based on the spectral data obtained from corn leaves polluted by Cu and Pb, the spectra were divided into four characteristic regions. Variational mode decomposition (VMD) was used to decompose the first-order differential spectrum, and the characteristic analysis was transformed from the spectral domain to the frequency domain. Each modal component was processed separately using principal components analysis (PCA) according to the different characteristic regions to obtain the main information on the pollution characteristics, and then a two-dimensional space was constructed to identify the differential characteristics of corn under Cu and Pb stress visually. Finally, the support vector machine (SVM) classifier was used to get the classification line model to distinguish Cu and Pb pollution. This method was named VMD-PCA-SVM. The results show that the method can highlight the spectral response characteristics of heavy metal pollution, which is expected to guide the rapid and non-destructive identification of heavy metal pollution in crops and the formulation of remediation strategies.