Abstract: In order to improve the practical training quality of students majoring in mechanical and electronics, an industrial robot sorting experiment platform was designed based on machine vision. The platform consists of four subsystems, including computer system, machine vision, robot and pneumatic system, which involve key technologies such as image target recognition and positioning, transformation of image coordinate system and robot coordinate system, and design of sucker control system. Six groups of innovative training projects are also designed. Through this experimental platform, students can learn image acquisition and processing, robot operation and upper computer software development. The teaching experiments with 30 students have attended in the last three sessions have proved that the practical training system plays an important role in cultivating students' learning interest, innovation ability and practical ability.