Research on Logistics Package Detection System Based on Lightweight MobileE-YOLOv8

In order to improve the intelligent sorting and transportation capabilities in warehousing and logistics, and solve the problem of difficult algorithm deployment on mobile devices. This paper proposes a lightweight YOLOv8 network that can quickly realize express parcel detection and sorting. First, the MobileNetv3 network is used to replace the backbone network of YOLOv8. Secondly, in order to improve the feature extraction ability of the network, the neck network is optimized, and the C2f fusion attention mechanism ECA is used to construct a feature extraction module, which effectively solves the problem of small defect features being difficult to extract under complex backgrounds. According to experimental verification, the MobileE-YOLOv8n model constructed in this paper has a 60.4% reduction in parameters compared to the YOLOv8n model, and the detection speed of the source code has been greatly improved. The amount of computation has been reduced by 65.4%. In addition, the size of the weight file is only 2.7M, which is 57.1% less than the source code, and ([email protected]) has only decreased by 3.4 % compared to the source code. The network constructed in this study is characterized by fast recognition speed and few network parameters. It can complete real-time target detection on embedded devices and can provide key technologies for the express parcel detection industry.