Abstract: In order to promote the inspection accuracy of cigarette seam impurities, small pieces of tobacco, dried glue and others, an inspection method based on machine vision was proposed. High-speed cameras were installed near the inspection drum and rejection drum in the cigarette maker separately to capture the images of each single cigarette from different positions and angles. Mean filtering and gray gradient calculation methods were used to determine the edges of cigarette seam, and then to determine whether there existed any impurities within the area enclosed by the edges. The cigarette with seam impurity was rejected through a rejection valve. The system was tested on cigarettes of two different specifications in Hangzhou Cigarette Factory. The results showed that under normal working conditions, the rejection accuracy for cigarettes with seam impurity was ≥99.5%, and the miss rate was < 1%; under abnormal working conditions, the rejection accuracy was ≥99.0%, the system showed high resistance to interference, the inspection time was ≤6 ms. This technology provides a support for improving the appearance quality of cigarettes.