In order to solve inner packaging caused by food packaging machine faults, a food inner packaging detection device was designed based on the machine vision. The device was mainly composed of image acquisition, image processing system, judging and eliminating system. Depending on ultra-small CCD cameras to detect inner liner, at the same time the defective inner liner was determined and rejected via PLC program. By adopting different defective inner liner and FLUKE thermal imager, the detection rates of the machine vision and the stability of the equipment were tested. The results showed that the detection accuracy of the machine vision based device for defective inner liner reached 95%. After development, the defective cigarette packs on white shift, eve shift and night shift reduced by 13.3, 8.4 and 8.0 packs per month, respectively, and all instruments met the requirements of long-term stable operation.