To select the best optimization model on parameter extraction in Lorentzian Brillouin scattering spectrum and improve the accuracy of the temperature or strain measurement, taking the objective functions based on linear and nonlinear least-squares fits as the study objects, the influence of the optimization model on the accuracy of the parameter extraction was systematically investigated. Brillouin scattering spectrum signals with different signal-to-noise ratios, scanned frequency intervals, g0, vB and ΔvB were numerically generated. The accuracy and computing time of parameter extraction and temperature/strain measurement of the linear model were compared to that of the nonlinear model. The results reveal that, if the signal is ideal and free from noise, the two models are free of any errors. The errors of the two models will increase with increasing in noise level/scanned frequency interval. However, the error of the nonlinear model is appreciably lower than that of the linear one. In engineering practice the nonlinear model should be selected. The analysis is validated with the experimental Brillouin scattering spectrum signals.