In order to improve the detection capability and accuracy of laser-induced breakdown spectroscopy (LIBS), the univariate and multivariate method [multiple linear regression (MLR)] are adopted to make quantitative analysis on Cr contained in samples respectively. Univariate analysis has been made by use of two characteristic spectral lines CrI: 425.435 nm and CrI: 427.48 nm, the limits of detection are obtained under the condition of five laser energies. The results indicate that the analytical results of CrI: 425.435 nm are better than CrI: 427.48 nm, the best limit of detection of Cr is achieved 5.8 mg/g. The linear correlation between the prediction and actual values of concentration of samples is discussed using MLR, the correlation coefficient is increasing to above 0.99 from 0.98 compared with univariate method. Using the approach of leave-one-cross-validation, the relative errors of prediction (REP) of univariate analysis of CrI: 425.435 nm and CrI: 427.48 nm are 6.73％ and 7.59％ respectively, while the REP of MLR is reduced to 4.66％ . The results indicate that LIBS combined with MLR can improve the accuracy of prediction on concentration of samples.