Abstract: Affected by the diversity of seed shape and volume, the disk-based seed arrangement is prone to missing seeding, multi-seeding and even cavitation during seeding. The uneven seeding directly affects the quality of crop production and seeding efficiency. In this paper, a precision plant-seeding method based on machine vision was proposed for different types of major cash crops such as corn, soybean and plateau summer vegetable in Gansu province. Firstly, The CCD module with light source is adopted in the seed waiting area to obtain the distribution and quantity of seeds, and to solve the problem of target blurring caused by uneven light exposure. Then the OTSU image segmentation method was performed to obtain seed centres and count them. Finally, decisions are made according to the counts, with replenishment for omissions and thinning for overplanted seeds. The experimental results of corn, soybean, and plateau summer vegetable seeds show that the seeding accuracy of this method (taking 2 seeds per seed pit as an example) is 99.46%, 99.85%, and 97.78% respectively, which is improved compared to the traditional seeding method.