Calibration method has an important impact on the accuracy of quantitative analysis of laser-induced breakdown spectroscopy (LIBS). In the process of using LIBS to detect copper in soil samples, different calibration curves are used in order to do quantitative analysis, including exterior standard method and different internal standard methods. Using Voigt fitting function to the spectral line at Cu 324.754 nm and Ca 317.9 nm to obtain the net intensity values, the net intensity ratio between Cu and Ca is then sought as internal standard to get calibration curve, which has the correlation coefficient up to 0.998. Through the cyclic inversion, the average relative error is 2.28% and the volatility is very small, so this method has good stability. Using the ratio between the spectra line intensity of Cu 324.754 nm and the area from 323 nm to 326 nm to establish the calibration curve, the correlation coefficient is 0.997 and the average relative error by cyclic inversion is 2.23%, but the stability is slightly lower than the former one. Directly using the Cu concentration and spectra line intensity to establish the calibration curve, the correlation coefficient is 0.972 and the average relative error by cyclic inversion is 13%, showing large fluctuations and less reliability.