In this paper, confocal Raman spectroscopy was applied to detect the contents of lead chrome green as a heavy-metal stain illegally added in tea. Firstly, Raman spectra of five different concentrations of lead chrome green in tea infusion were acquired based on specific concentration method. The qualitative analysis of sample added with lead chrome green was achieved with comparing the Raman spectra of sample and standard substance. Four main Raman characteristic wavenumbers, 1 341, 1 451, 1 527 and 1 593 cm-1, were extracted for the qualitative identification of lead chrome green in tea. After spectral preprocessing of the raw Raman spectra, backward interval PLS (biPLS), competitive adaptive reweighted sampling (CARS) and successive projections algorithm (SPA) were combined to deeply mine the characteristic wavenumbers of lead chrome green in Raman spectra, and finally 14 characteristic wavenumbers were optimized. Partial least squares (PLS) and least square support vector machine (LS-SVM) were separately used to build the model based on the extracted 14 wavenumbers. As a result, these two models both had good robustness and high ability to predict and all the determination coefficient (R2) of calibration, validation and prediction were higher than 09, which proved the effectiveness of the extracted characteristic wavenumbers. Compared with the PLS model, the nonlinear model built by LS-SVM got a better result, R2 of prediction was 0964 and the root mean square error of prediction (RMSEP) was 0535. This study indicated that it is feasible to detect the contents of lead chrome green illegally added in tea based on confocal Raman spectroscopy combined with specific sample treatment and chemometrics methods. This study helped the valid supervision of food safety problem on lead chrome green illegally added in tea.