This study aimed to improve the measurement accuracy of the discrete element method (DEM) simulation parameters of sorghum seeds for enhanced model accuracy. We measured the intrinsic and contact parameters of sorghum seeds, extracted the outline of the seeds by 3D scanning, and used the multi-spherical particle model filling method in the EDEM (Version 2020) software to obtain the simulation model of the seeds. By simulating the free-fall, slope slip, slope rolling, and rolling experiments of sorghum seeds–photosensitive resin material, we calibrated the collision restitution coefficient of the sorghum seeds–photosensitive resin material to 0.690, static friction coefficient to 0.345, and rolling friction coefficient to 0.040. Through the steepest ascent search and central composite design experiments, we calibrated the collision restitution coefficient of sorghum seeds–sorghum seeds to 0.400, the static friction coefficient to 0.450, and the rolling friction coefficient to 0.043. The angle of repose of the parameter combination and the angle of repose of the physical experiment (33.82°) were verified, and the relative errors of the coefficients were 0.7%, 0.25%, and 0.3%, respectively, indicating that the established model had a high simulation accuracy and reliability. These results show that the established sorghum seeds model and the optimally calibrated parameter combination are effective for DEM studies on sorghum seeds.