On the tobacco processing transportation line, when disassembling the tobacco packets packed in cartons, some foreign matters such as paper tape will be left behind. These foreign matters directly affect the automatic production efficiency of tobacco and reduce the quality of tobaccos. Tobacco foreign body detection system is a kind of automatic equipment which uses machine vision technology to detect foreign body in tobacco packets. It is used in the tobacco conveyor belt, leaf and rewire line, which is used to detect and remove the tobacco packets in the mixed foreign body of the automated detection system. At present, the detection method based on color difference is difficult to detect sundries that are similar to the color of tobacco, so the improvement of equipment performance meets a bottleneck. This paper proposes a Bayes detection method, which can detect foreign bodies that cannot be detected by color detection. It is of great significance to improve the performance of the equipment. The field test shows that the algorithm is effective and has certain application value.