Abstract: Aiming at the complexity of irregular cigarette parcel stacking in tobacco logistics, a system which combined cigarette parcel checking and stacking was designed. The system was composed of a parcel stacking system and a machine vision-based identification system. The identification system processed the image of irregular cigarette parcel to obtain target's binary image, extracted the corner point features of cigarette parcel to conduct feature point matching, and rejected the erroneously matched points via modified RANSAC algorithm to identify cigarette parcels. Then cigarette parcels were stacked following the pre-layering algorithm of host computer. The results of on-site test showed that the accuracy of cigarette parcel matching was over 97%, the time for feature extraction was less than 360 ms, the total time for parcel matching was less than 1.6 s. This method provides a support for promoting the automation of tobacco logistics.