Cavity confinement was combined with traditional laser-induced breakdown spectroscopy (LIBS), using Ba Ⅱ 495.709 nm as the analysis line to improve the quantitative analysis and detection of Ba in the soil. A univariate calibration model based on the spectral peak integration, multivariate principal component regression (PCR), and artificial neural network (ANN) calibration model was established to quantify the metal Ba content in the soil. Compared to traditional LIBS, cavity-confinement LIBS (CC-LIBS) increased the spectral intensity and signal-to-noise ratio of the characteristic spectrum. When Ba was analyzed using the spectral peak integration method, CC-LIBS could improve the precision of univariate quantitative analysis compared to traditional LIBS. CC-LIBS combined with multivariate regression model PCR and ANN was used to improve the detection accuracy of LIBS and reduce the matrix effect in the soil on the content analysis.In addition, the correlation coefficient of calibration curve was improved from 0.63 to 0.84. The mean relative error (MRE) of the verification set was reduced from 47.52% to 23.44%, respectively. And the detection limit for Ba element was reduced from 64.73 to 37.86, respectively. CC-LIBS combined with multivariate regression model PCR and ANN was used to improve the detection accuracy of LIBS and reduce the matrix effect in the soil on the content analysis. The correlation coefficient of multivariate regression calibration curve were 0.941 and 0.999, respectively. And MREs of verification set are 9.93% and 5.35%, respectively. This research provides a new idea for the application of LIBS technology to soil quality testing.