Design and Implementation of an Integrated Cigarette Recognition System for Tobacco Logistics

This paper presents a deep learning-based cigarette recognition system that integrates image recognition with barcode scanning, enabling compatibility with various package forms. The system leverages the SkipNet architecture combined with the MobileNetV3 feature extraction module to ensure high efficiency and accuracy in cigarette sorting operations. By deploying the YOLOv5 object detection algorithm within a real-time environment and introducing a Coordinate Attention (CA) mechanism, the system achieves significant improvements in processing speed while maintaining recognition precision. Experimental results demonstrate that the system processes up to 11 packages per second with an accuracy rate meeting industrial application standard, providing an efficient and reliable solution for tobacco logistics centers.