The weighing accuracy of electronic balance is determined by its linearity. In this paper, the piecewise linear approximation, least square fitting, Newton interpolation and BP neural network methods are utilized to deal with the nonlinear calibration of electromagnetic force equilibrium sensor. The experimental results show that the piecewise linear approximation method works best. Thus we design a calibration platform based on the piecewise linear approximation method. And it is able to achieve real-time and online compensation when changes take place in the ambient or sensor's parameters.