Rapid Detection of Arsenic in Water Based on Hyperspectral Imaging

With the intensification of industrialization and urbanization, arsenic pollution in water bodies has become a serious global issue. A rapid and pollution-free method for detecting arsenic content in water bodies is of great significance for water protection. In this study, quantitative analysis of arsenic content in water was conducted using hyperspectral imaging technology combined with a carrier. The arsenic solution was divided into three concentration gradients: 0.96879.0369 mg/l, 9.961190.0331 mg/l, and 99.9591∼900.0494 mg/l, totaling 540 samples, to investigate the concentration-dependent effects. The influence of filter paper on arsenic content detection was studied by adding different amounts of arsenic solution onto filter paper. A particle swarm optimization-least squares support vector machine (PSO-LSSVM) model was established using different preprocessing methods to determine the optimal preprocessing method. The competitive adaptive reweighted sampling (CARS) algorithm was used to extract characteristic wavelengths. Three models (PSO-LSSVM, BPNN, and SSA-KELM) were established based on full wavelengths and characteristic wavelengths, respectively. The results indicate that the PSO-LSSVM model established through characteristic wavelengths has a good predictive effect on arsenic content among all models. Specifically, at low concentrations,
$\textbf{Rp}^{2}=0.9951,\textbf{RMSE=0.1992},\textbf{RPD =12.5988}$
; at medium concentrations,
$\mathbf{R p}^2=0.9957,\textbf{RMSE =1.6853},\textbf{RPD=12.9468}$
; at high concentrations,
$\textbf{R}_{{p}^{2}}=0.9962,\textbf{RMSE =14.3584},\textbf{RPD=14.6843}$
. Hyperspectral imaging technology combined with a carrier was found to enable rapid and accurate prediction of arsenic content in water, potentially serving as an attractive alternative for online analysis in the field of water quality testing.