Detection of heavy metals in vegetable soil based on THz spectroscopy

Abstract

Heavy metal pollution in soil endangers food safety and human health. Thus, it is important to study accurate and rapid detection methods. Here, an efficient nondestructive detection method for mercury (Hg), cadmium (Cd) and copper (Cu) in soils was studied by terahertz (THz) spectroscopy. First, regression equations were established between heavy metal contents and absorption coefficients at the selected frequency points. Then, the pollution type and pollution level of the soils containing three heavy metals were detected at the same time. Reference blank soil was also tested. Probabilistic neural network (PNN) and random forest (RF) models verified the effects of qualitative detection. Next, the contents of the three heavy metals in soils were predicted simultaneously by a backpropagation neural network (BPNN) and an extreme learning machine (ELM). The results showed that the absorption coefficients increased regularly in the THz spectral range from 0.05 THz to 0.7 THz. The average detection result of the PNN model was better than that of RF. The average detection accuracy for heavy metal pollution level and type were all higher than 95%. In addition, the prediction results of heavy metal content showed that BPNN model has better prediction performance. The optimal decision coefficients (DC) of BPNN model for soils containing three heavy metals were 0.95, 0.99 and 0.98, respectively, and their corresponding root mean square errors (RMSE) were 0.37, 0.02 and 2.62, respectively. The results proved that THz spectroscopy has good qualitative and quantitative detection ability for soils contaminated with Hg, Cd and Cu, which could bring new opportunities for detection of heavy metal pollutants in soil.