Magnetic-type seeder is a precision metering device, particularly suitable for small seeds. A visual inspection system for precision performance of magnetic-type seeder was established based on machine vision and image processing technology. By using pre-treatment techniques including image binarization and image filtering, image quality was enhanced effectively. As the gray-scale value of the seeds coated with magnetic powder is very close to the electromagnet, the method of seed feature extraction based on morphological image processing is proposed and performance testing model of precision metering seed is put forward. The results of comparison between machine vision and manual detection showed that the relative error of preciseness was less than 3% and coefficient of variation and standard deviation were less than 5%, which indicated the system is of high accuracy when used in real-time detection.