In order to solve the problem of missing seeding caused by small impurities or damaged seeds blocking suction holes under the action of adsorption force in the process of filling, clearing and carrying seeds, the unloading mechanism was optimized. The installation position of seed unloading mechanism was improved to ensure that the seed metering device could adsorb seeds well in the seed-carrying area, prevent flying seeds, and reduce collision and bounce. The probability of seed falling off at the end of the air chamber in the seed-carrying area was reduced by 1.67% compared with that before the improvement. A gear tooth surface curve suitable for the coordination between the unloading wheel and the seed plate was deduced. By means of ADAMS simulation, three indexes of meshing force, radial force and axial force were extracted, and the rationality of the unloading gear tooth design was verified by simulation. The curve equation can be used to design unloading wheels with different kinds of plate and suction holes, which made the meshing stable and reliable and had good versatility. A comprehensive test was carried out with the unloading mechanism, forward speed and wind pressure as factors. By analyzing the significant difference between unloading mechanism and wind pressure at different speeds and the overall variance of the experiment, the key factors affecting the qualified index, multiple index and missing index were determined. The optimized new unloading mechanism was selected for regression analysis, through regression equation, the optimum operating parameters of the designed metering device at operating speeds of 10km/h, 12km/h and 14km/h were obtained, and the experimental verification was carried out. The results showed that the new unloading mechanism can effectively improve the qualified index and reduce the missing index of high-speed operation. When the operating speed was 10~14km/h and the wind pressure was -3.81~-3.43kPa, the qualified index can reach more than 96.8%, the missing index was not more than 2.0%, and the multiple index was not more than 1.2%. All the indexes were better than the requirements of national standards and can achieve effective seeding.