Automated cargo recognition and localization system in smart warehousing based on IoT and improved YOLOv8
Abstract
This paper proposes an automatic identification and positioning system for smart warehouse goods based on the Industrial Internet of Things (IIoT) and an improved YOLOv8 architecture. By introducing a cross-channel attention mechanism into the network and optimizing the intersection-union bounding box constraint matrix, the stability of feature extraction is improved. A mapping process from 2D pixels to 3D physical space is established based on edge heterogeneous computing units combined with dynamic filtering equations. Test results show that the architecture achieves an average accuracy of 92.4% under specific hardware constraints, with an end-to-end integrated inference latency of 15.8 milliseconds, communication efficiency maintained at 95%, and high accuracy in 3D spatial positioning. This provides quantitative analysis and system deployment reference for the application of visual perception technology in logistics hub platforms.