Abstract: To promote the automation level, the process of empty cartons handling was upgraded with an unstacking system based on a high-precision camera. A YOLO v8-CBAM-C2fDynamic model was established by adding a CBAM (Convolutional Block Attention Module) and a C2fDynamic (Dynamic Convolution Module) to the YOLO v8 model. The template matching model along with the visual camera is used to identify the presence of empty cartons on the pallet in the unstacking station, the YOLO v8-CBAM-C2fDynamic model is used to identify the orientation and positions of the empty carton group and guide the robot to accurately pick up the empty carton group and transfer it to the film wrapper removal station. The empty cartons were transported by a conveyor belt to a carton filler after the film wrapper was removed from the carton group. The designed system was tested with the empty hard cartons of "Yellow Crane Tower 1916" cigarette brands in Wuhan Cigarette Factory. The results showed that: 1) YOLO v8-CBAM-C2fDynamic model could accurately and quickly identify the orientation and positions of empty cartons, with mAP (Mean Average Precision) of 0.904 and FPS (Frames Per Second) of about 120 frames per second. 2) Compared to manual feeding, the designed system reduced the number of workers by 3 per shift. 3) Compared with the template matching model, the success rate of the designed system increased from 67% to 100%, and the unstacking efficiency increased from 40 to 64 pieces/min, meeting the feeding requirement of a carton filler. This technology supports the promotion of the automation level of cigarette cartoning.