Abstract: In order to further improve the intelligent operation level and picking efficiency of picking robots in China, optimization research was carried out based on computer vision application technology. Taking the structure group of the picking robot as the design basis, the mathematical model of the visual recognition processing and control of the picking robot was established by using the core concept of visual control, the appropriate picking path planning and the system picking state output design were implemented, and the visual picking verification test was carried. The results showed that the optimization of the picking robot system based on computer vision was correct and feasible, and the overall picking efficiency of the whole machine could reach 94.61%, the recognition accuracy and control accuracy of the system were increased by 6.12% and 5.25% respectively; the robot picking success rate could be increased to 95.98% due to the accuracy of the computer vision processing technology and the flexible coordination of the picking execution components. The effective application of the computer vision processing technology to the design and improvement of the picking robot would provide innovative and reference ideas for the development and research of similar agricultural picking and harvesting equipment, with good promotion value.