As important factors in discrete elements, the physical parameters of watermelon seeds play a pivotal role in discrete element method. To obtain the discrete element parameters of watermelon seeds and improve the accuracy of the discrete element model, through a combination of actual and simulation tests, this research has calibrated the seeds’ physical and contact parameters with the seed metering device. Employing the Plackett-Burman experiment, this study has identified three critical factors affecting the stacking angle: the static and rolling friction coefficients between seeds, and the collision recovery coefficient between seeds and plexiglass. Using the steepest-climbing design and Box-Behnken response surface analysis, this research has optimized these factors to values of 0.716, 0.051, and 0.787, achieving a calibration error of just 2.60%. Verification with an air suction precision seed metering device confirmed the parameters’ accuracy, with relative errors below 7.65%. The discrepancy between the simulation and actual test results, as measured by the qualified index error, is successfully reduced to below 4.38%. This study thus establishes a solid foundation for the structural optimization of air suction precision watermelon seed metering devices.