The work aims to design a visual feeding system based on machine vision to improve the automation of the feeding and delivery process in the packaging production line. Firstly, the principle of camera imaging was introduced. On this basis, the visual calibration process, including rigid body transformation, perspective projection, image discretization, etc., was discussed in detail. The structure of visual feeding control system which was mainly composed of Ipc, software platform and human-computer interaction interface was discussed. Ipc could realize the distributed control and centralized management of vision acquisition module, robot control module and sensor signal acquisition module. The effectiveness of the control system was verified by experiment. The experiment result showed that the maximum positioning error was only 0.7 mm and the positioning accuracy was relatively high. The success rate of packing could reach 99.5% or above, and there would be no mis-grasping or missing grasping. The system responded quickly. The visual feeding control system of packaging production line has the advantages of fast response, high positioning accuracy and high success rate of packing. It meets the requirements of high speed and high precision in packaging process.