The work aims to solve the problems of 'excessive cigarettes', 'lack of cigarettes', 'wrong cigarettes', etc. in the cigarette sorting system due to the large workload of cigarette sorting in tobacco distribution centers, so as to improve the sorting efficiency. A cigarette recognition system based on template matching to extract the target cigarette image was designed. In order to improve the image quality, the image shooting platform adopted a semi-closed tunnel device. After the image was preprocessed, the target cigarette outline was extracted accurately and completely by determining the scanning template size, fixing the starting point of image to be scanned and applying the secondary matching method of coarse matching and rematching. Then, the image features of the cigarette image were extracted through the Harris corner detection algorithm to establish a database to identify and correct errors based on RBM model. The experiment in the distribution center showed that the error correction rate of cigarette recognition system was about 99.9%, with a false alarm rate less than or equal to 1/30 000 on a conveyor belt with a speed of 1.5 m/s, and the system was stable. The cigarette error correction system solves the cigarette sorting errors well, improves the cigarette delivery efficiency, and meets the error correction requirements of the tobacco logistics distribution center.