Abstract: To realize the rapid and nondestructive measurement of moisture content in corn, the dielectric constant (?′) and the dielectric loss (?″) of corn kernels of different moisture content under different temperature during hot air drying process were measured at an excitation voltage of 500 mV and frequency range of 1 kHz to 5.462 MHz with a precise impedance analyzer and a self-made dielectric parameter measuring sensor, and the mechanism of change in dielectric parameters during drying process was studied and analyzed; meanwhile, a support vector machine-based moisture content prediction model was established by incorporating double dielectric parameters, and a model verification experiment was carried out. It was shown by the experiment results that the dielectric constant (?′) of the corn kernels in the different moisture content decreased with the increase of measurement frequency; the dielectric loss (?″) of the corn kernels in the low moisture content increased first and then decreased with the increase of frequency, with an obvious relaxed peak observed within the frequency range of 1 kHz-1 MHz, while for the corn kernels in the high moisture content, the dielectric loss (?″) decreased with the increase of measurement frequency; in addition, with the increase of frequency, the tangent value of the dielectric loss angle increased first and then decreased for corn kernels in the low moisture content and decreased for the corn kernels in the high moisture content. At the measurement frequency of 5.462 MHz, the dielectric loss angle of the corn kernels decreased with the decrease of the wet-basis moisture content, the first derivative of the dielectric loss angle versus the moisture content dtanδ /dM showed the obvious trend of the segmentation with the decrease of moisture content. Through the analysis of the double dielectric parameter spectrum, the frequency bands for modelling had been selected initially. A total of 31 dimensional variables, including the double dielectric parameters of 15 measured frequency points between 1.072 and 5.462 MHz and the temperature value T, were used as the input full variables of the support vector regression machine (SVR) model. Competitive adaptive reweighting algorithm (CARS), iterative retention information variable algorithm (IRIV) and CARS-IRIV joint algorithm were used to screen the feature variables, respectively. SVR models were established respectively with full variables and the feature variables screened by CARS, IRIV and CARS-IRIV. By optimizing the parameter c (penalty factor) and the parameter g (kernel function parameter) of the SVR model with the WOA (whale optimization algorithm), the result indicated that the optimal prediction performance could be witnessed in the CARS-IRIV-WOA-SVR model. The feature variables screened by CARS-IRIV were ?′3.854MHz, ?″3.854MHz, ?′5.462MHz, ?″5.462MHz and T. Determination coefficients of the prediction set, RP2, RMSEP (root mean square error of prediction) and the RPD (residual predictive deviation) were 0.998 4, 0.40% and 24.55, respectively. This study provided a new research idea and basic data for rapid nondestructive testing of maize moisture content based on double dielectric parameters and SVR, and it also laid a foundation for the development of intelligent grain moisture content rapid nondestructive testing instrument.