基于轻量卷积神经网络的智能电子秤设计  

A Study on the Design of the Intelligent Electronic Scale Based on the Lightweight Convolutional Neural Network

作　　者：郑冬 李萍萍 龚识懿 熊美玲 潘亚夫 ZHENG Dong;LI Pingping;GONG Shiyi;XIONG Meiling;PAN Yafu(Chongqing Electric Power College,Chongqing 400053,P.R.China)

机构地区：[1]重庆电力高等专科学校,重庆400053

出　　处：《重庆电力高等专科学校学报》2023年第5期18-21,共4页Journal of Chongqing Electric Power College

基　　金：重庆市九龙坡区科技局科学研究项目(C-KY202304)。

摘　　要：为解决商品交易过程中存在的效率性和准确性问题,运用OpenMV单目视觉和压力传感器设计了一款智能电子秤,并分别运用商品识别和质量检测进行了实验验证。在OpenMV单目视觉中融入轻量卷积神经网络模型,模拟不同光照下的香蕉、樱桃、石榴、草莓等水果进行了实验,其商品识别的准确率约为94.16%。此外,利用压力传感器对50 g、100 g、200 g等标准砝码进行质量测量,相对误差位于[0.05%,1.27%]之间,且误差与质量之间存在负相关。商品识别和质量检测实验结果呈现两者误差相对较小的特点,满足商品交易的实际需求,也直接反映出利用轻量卷积神经网络实现智能电子秤具有合理性。In order to solve the problems of efficiency and accuracy in the process of commodity transactions,this paper introduces an intelligent electronic scale designed by using the OpenMV monocular vision and pressure sensor,and the experimental verification is carried out by means of commodity identification and quality detection.The lightweight convolutional neural network model is integrated into the OpenMV monocular vision,and the experiment is implemented by simulating merchandises such as bananas,cherries,pomegranates,strawberries,etc.under different lighting conditions in supermarkets,and the accuracy of merchandise identification is about 94.16%.In addition,the weight measurement using standard weights such as pressure sensors of 50g,100g and 200g has relative errors between 0.05%and 1.27%,and there is a negative correlation between the errors and weights.The experimental results of commodity identification and quality detection show that the errors between the two are relatively small,which meets the actual needs of trade,and also directly reflects the rationality of using the lightweight convolutional neural network to create the intelligent electronic scale.

关 键 词：效率性 OpenMV单目视觉 轻量卷积神经网络 负相关 

分 类 号：TH715.193[机械工程—测试计量技术及仪器]