Rice is one of the most important food crops in China. Rice direct seeding is becoming more and more popular in China for its low-input demand. A centralized pneumatic metering system for rice direct seeding was involved, which distributed seeds to 33 rows through only one metering system. There were two main ways to adjust the seeding amount of a rice seeder. One was the effective length adjustment of seed meter and the other one was the adjustment of speed ratio of metering wheel and machine. Either way treated seeding rate of each turn (SRE) as a fixed value, and the accuracy of seeding rate adjustment was poor. Seed height (h), speed of the seed metering wheel (n) and seed moisture content were the key factors for SRE of the metering system. In order to improve the accuracy of seeding rater of the centralized pneumatic metering system, experiments were carried out through a metering device test-bed JPS-12, seed height (h) and speed of the seed metering wheel (n) under four different moisture contents of seeds were selected as the experimental factor while SRE was selected as the test index. The effect of seed height (h), speed of seed metering wheel (n) and seed moisture content on SRE was revealed by the analysis of the test data. The experiment showed that SRE of different seed moisture content was decreased with the increase of rotation speed of the metering wheel. When the height of the seed height was fixed, the two was inverse ratio. SRE was increased at first and then decreased with the increase of seed height, and the maximum of SRE appeared when the seed height was between 25.35cm and 31.55cm. Under the four different seed moisture contents, SRE from large to small were dry seed, seed that dried two days, seed that dried one day and wet seed, and SRE was decreased with the increase of seed moisture content. With two-regression analysis, the regression model of SRE with seed height and speed of the seed metering wheel under four different seed moisture contents was acquired, and a seeding amount control model based on the control of speed of the seed metering wheel was constructed. The test platform was built to test the seeding amount control model. The average error of the control model was 2.07%, and the actual sowing quantity variation coefficient was 2.59%. The verification results were basically consistent with the control model. The control model of rice seeding amount under the influence of multiple factors was established. The influence of seed height, speed of the seed metering wheel and seed moisture content on SRE was clarified, which could be used for the design and optimization of rice seeding amount control system.