Abstract: Abstract: Garlic is one of the most important economic crops widely planted with an annual planting area of 10 million mu in China. However, there is a relatively low mechanization level of garlic planting at present, particularly with high labor intensity and low efficiency. In this study, a garlic seed metering device with a double seed-filling chamber was designed for the better performance of mechanized sowing, including a high single seed rate, and a low missing rate of garlic. The device was composed mainly of a seed box (first seed-filling chamber), transmission shaft, transmission chain, sprocket, seed picking spoon, seed guard plate, second seed-filling chamber, and guiding port components. A systematic analysis was also made on the main structural parameters and working principles of device ware. An optimal arrangement was proposed for the series combination of the front and rear seeding spoon and the coordinated seeding ware. The speed of the first and second spoon was required to meet a reasonable seed takking speed ratio. The reason was that the number of garlic seeds was required not less than one in the second seed-filling room after the first spoon extracting garlic seed, in order to gain a low missing rate of garlic seeds. At the same time, a new mechanism was proposed to explain the decrease in the missing rate of garlic seed in the double seed-filling chamber. A discrete element software EDEM was used to analyze the influence of the speed of the first spoon and the thickness of the seed layer on the filling performance. The results show that the optimal seeding speed and thickness of the seed layer were beneficial to the filling of seeds. A kinetic model was established for the garlic seed filling and removal. The speed-time curve of garlic filling and removal was then determined in the second seed filling chamber. A piece of self-developed equipment was selected to verify the obtained data. A Box-Behnken central composite experiment was conducted to obtain the optimal working parameters for the garlic seed metering device with a double seed-filling chamber. A three-factor three-level orthogonal experiment was also carried out, where the first order spoon speed, seed speed ratio, and seed layer thickness were taken as the influence factors, while the single seed qualified rate and the missing rate of garlic as the response indexes. A mathematical regression model was then established for the test factors and evaluation indexes after each experiment using Design-Export 8.0.6 data analysis software, where the parameters were optimized. Specifically, an optimal combination was achieved, where the speed of the first seed taking scoop was 0.12 m/s, the seed speed ratio was 0.75, and the seed layer thickness ranged from 360 to 390 mm. An indoor verification test was also carried out to verify. In this case, the single seed qualified rate, the missing rate, and the replay rate were 95.38%, 1.18%, and 3.44%, respectively, and the coefficients of variation were 0.32%, 6.11%, and 4.15%, respectively, particularly, the relative error between the verified and predicted value was less than 5%. Furthermore, the mechanical damage rate of garlic seed was 0.52%, indicating the better adaptability to 3 types of garlic Ⅱ grade seeds. The field test demonstrated that the data was consistent with the optimized one, indicating better performance than the previous device. This finding can provide sound guidance and reference to deal with the leakage of garlic planter.