Abstract: Accurate estimation of model parameters gives a better prediction basis for model validation and application. In tomato model, non-linear least square method(NLSM) was applied to calibrate growth parameters by fitting multi-stages tomato morphology data based on the description of their topological structure. The weight values in NLSM were adjusted automatically according to target data, which allowed the measured data with different dimensions or ranges to be treated at the same level, and thus the accuracy and stability of algorithm were improved accordingly. Derivative was replaced by the difference quotient in NLSM which further improved the algorithm speed. The check results of the optimum grade about the data fitting indicated that model output fitted well the trend of observed data, which verified that estimated parameters were credible, and model was valid on a certain scale. The practice proved that the multi-target fitting method is an effective and powerful tool applied to the crop model parameter calibration.