The water-oil relative permeability curve is mainly obtained from linear displacement experiments. Few radial displacement experiments have been carried out. In the process of linear displacement experiments, the flow properties of the water-oil two phase are linear. Nevertheless, it is radial near the bottom holes of an actual reservoir. With regard to both kinds of displacement experiments, the flow characteristics are various, which may result in great deviation to apply the linear calculation. theory of the relative permeability curve to an actual reservoir. As a result of the above-mentioned problems, on the basis of radial displacement experiments, using the Levenberg-Marquardt algorithm for automatic history matching, this paper performs optimization of production performance and relative permeability representation models. Finally, a novel numerical inversion method for the radial water-oil relative permeability curve is established. A test based on the basic data of a radial laboratory displacement experiment is performed to verify the effect of the proposed method. The results show that oil relative permeability is more sensitive to the cumulative production data, while water relative permeability is more sensitive to the bottomhole pressure data of the producers. It also indicates that the cubic B-spline model (CBM) is far more general and flexible and has the advantage of local fitting compared to the power law model (PM). In addition, the numerical inversion method proposed for the radial water oil relative permeability curve is reliable and can meet the engineering requirement, which provides a basic calculation theory for the estimation of the water-oil relative permeability curve.