基于边缘计算的超市水果智能电子秤

Intelligent Electronic Scale for Supermarket Fruit Based on Edge Computing

摘要: 水果种类多，种间相似大，需人工辅助识别后称重销售，电子秤响应时间受网络影响大，智能化程度低导致用户体验差。文章采用基于边缘计算的YOLOv4-Tiny算法和高精度传感器，实现了边缘式超市水果智能电子秤。首先，采用双孔平行梁式传感器作为主体搭建电子秤；其次，对YOLOv4-Tiny添加注意力机制CBAM，提高相似水果的识别率，并将模型部署于边缘设备Jetson Nano上；最后，建立了基于本地边缘计算与远程存储服务一体化的超市水果管理系统，实现数据可追溯。实验结果表明：该智能电子秤能够实时在线识别货物种类，并将货物名称、货物重量、称重时间和货物价格等信息存储在远程服务端，电子秤还支持语音提示、人机交互、在线支付和凭证自动打印等功能。系统响应速度快，用户交互体验强，可为有实时检测需求的应用场景提供借鉴。

Abstract: Due to the variety and similarity of fruits, manual recognition is required for weighing. The response time of electronic scales is greatly affected by the network, and the low level of intelligence leads to poor user experience. In this paper, YOLOv4-Tiny algorithm based on edge computing and high-precision sensors were used to realize the supermarket fruit intelligent electronic scale. Firstly, a dual hole parallel beam sensor was used as the main body to build an electronic scale; Secondly, attention mechanism CBAM was optimized the model of YOLOv4-Tiny to improve the recognition rate of similar fruits, and the model was deployed on the edge device of Jetson Nano; Finally, a supermarket fruit management system based on the integration of local edge computing and remote storage services was established to achieve data traceability. The experimental results show that the intelligent electronic scale can recognize the type of goods in real-time online, and store information such as the name, weight, weighing time, and price of the goods on the remote server. The electronic scale also supports functions such as voice prompts, human-computer interaction, online payment, and automatic voucher printing. The system has fast response speed and strong user interaction experience, which can provide reference for application scenarios with real-time detection needs.