Cigarette detection in express packages for terahertz imaging based on YOLOv8

With the development of the express delivery industry, many fake sellers use express delivery to sell fake cigarettes, which brings new challenges to the supervision of tobacco monopoly. In order to improve the work efficiency of tobacco law enforcement officers, this paper uses terahertz imaging technology to detect cigarettes in packages. Due to the low resolution and serious blur of terahertz image, the detection accuracy is seriously affected. In this paper, terahertz scanning imaging system is used to establish a terahertz cigarette datasets. Based on YOLOv8n, bi-directional feature pyramid network (BiFPN ) module is introduced to reduce the loss caused by the network feature processing by fusing multi-scale information. In order to improve the accuracy and efficiency of bounding box regression, Efficient IOU (EIOU) Loss is used. We tested our method on the datasets. The experimental results show that Precision (P) and the mean Average Precision (mAP) of the proposed method applied in cigarettes detection task are 96.1% and 97.3%. Compared with the baseline YOLOv8n, the results show that our proposed method improves the average accuracy by 0.6% ([email protected]).