Abstract: In order to promote inspection accuracy and reduce labor intensity, an online cigarette appearance inspection system based on machine vision was designed. The system used a CCD camera to acquire and pre-process cigarette images, the appearance of cigarettes was analyzed by an established two-stage inspection model. At the first stage, a maximum contour area method was adopted to identify the cigarettes with obvious appearance defects; at the second stage, a template method was adopted to identify the cigarettes with minor defects, and then the unqualified cigarettes were rejected. Two hundred cigarettes of brand "Liqun (new)" were sampled and tested in Ningbo Cigarette Factory, the results showed that the two-stage model accurately identified all sampled cigarettes. It identified 50 qualified cigarettes and 49 unqualified cigarettes in the test set of 100 cigarettes, the inspection accuracy reached 98%. This technology provides a support for promoting the automation level of cigarette makers.