Aiming at the requirements of machine vision industrial application and talents training in colleges and universities, and analyzing the problems existed in machine vision curriculum teaching, the authors construct a machine vision curriculum system that combines online and offline teaching together, which is proved to be oriented, staged and modular. At the basic stage, corresponding to each basic experiment, the knowledge points are divided to deepen theory understanding; at the innovation stage, the course content, competition and scientific research are combined to improve students’ ability of innovation and teamwork. The machine vision curriculum construction scheme provides a new teaching method and thought for the interdisciplinary subject with high practical requirements.