The work aims to design a packaging quality detection system based on machine vision to improve the accuracy of packaging defect detection. The hardware structure of machine vision system was introduced, including carrier platform, industrial camera, light room, computer, etc. On this basis, the image processing algorithm was described in detail with the detection of cigarette packaging as the main research object. First, the median filter was used to eliminate the original image noise. Then, the image edge was sharpened based on Canny operator. Finally, image registration was used to judge whether there ware defects in cigarette packaging. The validity of the system was verified by experiments, and the results showed that the recognition rate of the system to qualified samples could reach 100%. The recognition rate of unqualified samples could also reach 98.67%. The overall recognition rate was 99.33%. The machine vision system has the characteristics of high identification accuracy and stable performance. The image processing algorithm can accurately distinguish whether a cigarette has defects and has practical application value.