A seeding accuracy test experiment based on soil substrate was investigated. The moving trolley carried the sowing seeders. Meanwhile, the machine vision system sequential images dynamically were collected. Using encoder and PLC etc., it realized to control camera to collect sequential images through detecting the motion distance of trolley, which realized hardware splicing of the sequential images. In addition of splicing algorithm of image processing, the identification of seeds and the total splicing of the sequential images were completed. Experimental results showed that the matching rate of seed spacing was above 85%, and the accuracy rate of seed spacing was more than 70%. The experiment was repeated many times. The absolute error of the seeding accuracy data was within 4%.