Moisture content is a major index in the healthy growth of crops. It is beneficial to water and fertilizer management when the crop moisture content is detected timely. The dielectric properties (relative dielectric constant ε′ and dielectric loss factor ε″) of 280 pieces of corn leaves with different moisture contents were measured with a self-made clamping capacitor and an LCR measuring instrument at 36 discrete frequencies over the frequency range of 0.06~200kHz and the moisture content of the corn leaves were measured by drying weight method. To obtain the moisture content of corn leaves, linear regression methods (the combination of SWR and MLR) and nonlinear regression methods (SPA and SVR) were used to establish models to get the relationship between the moisture content and dielectric parameters (ε′, ε″ and the combination of ε′ and ε″), and the leave one out cross validation (LOOCV) was used to select the best models. The results showed that contrasted with the linear regression method, the nonlinear regression method had better predictive ability. The highest coefficient of determination (0.804) and the lowest root mean square error (0.0176) were obtained by using the nonlinear regression model with the variable in the combination of ε′ and ε″, which simplified the model with variables reduced from 72 to 10 and eliminated the overlap variables, and the complexity of the model was decreased effectively. The study indicated that it was feasible to detect the corn leaf moisture content non-destructively, and the results provided a credible method for rapid non-destructive detection of physiology information in crops.