To meet the needs of screening and rejecting defective products in the production and processing of green beans, an automated online screening system was designed based on the machine vision technology. The combing mechanism in the hardware device of the system arranged the green bean products on the conveyor belt in an orderly manner, making it a waterfall-style single-row free-fall motion. In the process of falling, the upper computer controlled two linear arrays of CCDs installed in high and low positions to obtain single-sided images of each green bean from the front and rear directions. The software system could process the collected green bean images in real-time, extract the proportion of black pixels of green beans and analyze the circularity through algorithms, extract the color and shape features of green beans, and encode the positions of defective green beans. And then the detected information was sent to the lower single-chip microcomputer in real-time, and the single-chip microcomputer controlled the multi-channel high-pressure air nozzle to complete the removal action. Thereafter, the online screening and removal of defective green beans were realized. The experimental results showed that the detection efficiency of the system was high, and the detection accuracy could reach more than 98%, which could effectively meet the needs of automatic and intelligent production of green beans, showing great value of promotion and application.