Research on key technologies for multiview fusion intelligent scanning in automated logistics lines for original tobacco packaging

The National Unified Tobacco Leaf Production and Operation Management Platform (the "Tobacco Leaf Platform") is a critical infrastructure for driving the digital transformation of tobacco leaf production and management. Implementing the "Two Strikes and Eight Sweeps" operation has significantly enhanced the quality and efficiency of tobacco package warehousing and logistics planning and control. However, in the context of automated scanning on the logistics line for tobacco packages, complex factors such as distortions in the appearance of tobacco packs and labels, loss of detail, blurred contours, and variations in illumination frequently result in missed scans, thereby hindering the promotion and practical application of automatic scanning technology in production. To this end, this paper proposes an intelligent solution: First, a novel dataset comprising thousands of images of tobacco packs, labels, and QR codes was constructed, filling a gap in the industry's data resources. Subsequently, a multi-view fusion image restoration method was developed to integrate multi-view information to enhance image quality while suppressing distortion and blurring effects. Finally, an online multi-scale object detection model was developed to achieve real-time and accurate detection and recognition of pack labels under complex lighting conditions. Experimental results demonstrate that when used in conjunction with the standardized dual-hole labels, the developed system exhibits excellent performance in the tobacco pack scanning scenarios on automated logistics lines, with a significantly reduced miss-scan rate. This study not only provides support for upgrading the tobacco leaf platform but also offers valuable theoretical and practical insights into automation in industrial logistics scenarios.