Manual supplemental seeding in plug seedling production is a labor intensive and very tedious task. This paper studied and developed an automatic supplemental seeding device using machine vision technique for corrections of seeding errors of plug seeder, and replacement of the manual reseeding. The device includes five units: sweet corn seed detection, seed releasing, seed picking up, and plug tray conveying. A voting mechanism-based sweet corn seed detection method was proposed for detecting the missing seeds in a moving plug tray. The seed releasing cells and the suction nozzle for picking a seed up were fabricated using 3D printing technique. The proposed device was placed into the practical growing seedling pipeline to test its reseeding performance. The results showed that the sweet corn seed detection accuracy was about 99%, the empty cell rate decreased from 9.61% to 0.38% (after reseeding) on average, and the reseeding rate was about 221 trays per hour when the miss-seeding rate was under 5%. This proved that the proposed supplemental seeding device has great significance for increasing the seeding precision of sweet corn, and improving the efficiency of the practical seedling.