The olivine on the lunar surface is a diagnostic product of the moon′s formation and geologic evolution. Its abundance distribution is a hot spot in lunar exploration. It is possible to inverse the global lunar olivine abundance using the obtained reflected spectra. In this paper, 51 groups spectral data of lunar soils produced by the lunar soil characterization consortium (LSCC) and its corresponding olivine abundance are used to establish four regression models. The remaining 6 groups data are used to validate the models. Based on the calculation of coefficients, combined with validating LSCC data, the standard deviation, correlation coefficient and scatter plots of different models are compared, and the best optimized formulation statistically with the relationship between spectrum and olivine abundance is chosen. By using the Clementine UV/VIS/NIR data and the model, the global lunar olivine abundances are inversed and mapped. By comparing with the measured olivine abundance of the Apollo sampled soils, it is proved that the result is reliable.