Abstract: Garlic is one of the most important crops with a pungent flavor and economic value in the fields of food and medicine, due to its rich nutritional components and bioactive substances. Thus, garlic has been widely cultivated in China, accounting for more than 80% of the total cultivation area worldwide. However, manual sowing of garlic cultivation cannot fully meet the requirements of large-scale production in recent years, particularly due to high labor intensity and costs, low efficiency, and low sowing quality. There is an urgent demand for high-speed and efficient sowing for large-scale cultivation in modern agriculture. Among them, single-seed picking can dominate the overall quality of mechanized garlic sowing. Existing garlic single-seed metering devices are also confined to the high multi-seed rate, empty seed rate, and low single-seed rate. In this study, a flexible opening-closing metering device was designed for a single garlic seed, according to the concept of “retain one from multiple seeds”. Theoretical analysis was conducted on the kinematics and dynamics of key components and garlic during seed picking. Key influencing factors on its seed picking performance were determined after structural optimization. A discrete element method (DEM)-multibody dynamics coupling model was established to simulate the metering device. Box-Behnken test was conducted to investigate the influence of structural parameters of the seed spoon on seed picking performance. The initial angle of the openable sidewall, the vertical height of the openable sidewall, and the radius of the hemispherical groove were employed as test factors, while the single seed rate, empty seed rate, and multi-seed rate were taken as test indices. Regression models of indices and factors were established to verify the simulation. The optimal parameters were determined after verification. Simulation test results indicate that when the initial angle of openable sidewall was 30°, vertical height was 17.5 mm, and radius of hemispherical groove was 11 mm, the best performance was achieved in the single seed rate of 95.69%, the empty seed rate of 1.83%, the muti-seed rate of 2.48%, and the error between simulation and theoretical calculation of 0.33 percentage points, indicating the reliable parameter after optimization. The coupled simulation was then analyzed to investigate the single-seed picking of the device under optimal parameters. The seed picking and seed cleaning of the device were determined according to the velocity, movement trajectory, and force conditions of garlic during simulation. The bench test revealed that the better performance was obtained with an average single seed rate of 95.04%, an empty seed rate of 1.49%, and a multi-seed rate of 3.47%. The error of the single seed rate between the bench test and simulation was 0.65 percentage points. The seed cleaning captured by a high-speed camera was consistent with the simulation, indicating the reliability of the model. Bench tests were also conducted to investigate the effect of linear velocity on the seed picking performance of the spoon, particularly at linear velocities of 0.05~0.40 m/s. The single-seed rate exceeded 85% at all linear velocities, indicating suitability for different linear velocities and high stability in single-seed picking. The single seed rate shared a trend of first increasing and then decreasing, as the linear velocity of the seed spoon increased. The single seed rate of different garlic sizes exceeded 93%, indicating the strongly flexible application of the seed spoon to garlic size. These findings can provide a valuable reference to optimize the metering device for a single garlic seed.