中国烟草学报2024，Vol.30Issue(4)：113-122,10.DOI:10.16472/j.chinatobacco.2022.T0173
基于深度学习技术的卷烟陈列智能化督导的研究与应用
Research and application of intelligent supervision for cigarette display based on deep learning technology
摘要
Abstract
To solve the problems of long time consumption and low efficiency in the supervision of terminal cigarette display by customer managers of the tobacco company,this paper designed an intelligent cigarette display supervision system(Cigarette Display Intelligent Guidance System,CDIGS)based on deep learning technology.The CDIGS system uses a cigarette product recognition algorithm based on feature matching and a cascade of classification algorithms.It recognizes cigarette display images taken by mobile terminals,automatically obtains the retailer's cigarette display,warning label placement information,and data related to cigarette display compliance rate,display specification count,and key brand shelving rate.By comparing with the preset standard model of cigarette display,the system provides real-time secondary display suggestions to retailers.Taking 516 cigarette retail stores in Baise City,Guangxi as an example,the CDIGS system was validated.The results show that:(1)The picture re-photography recognition model has an identification rate of 79%,effectively improving the authenticity of captured images;(2)The accuracy rates of the warning label detection,cigarette product detection,and cigarette brand recognition models all exceed 97%,accurately extracting cigarette information from images;(3)After a brief introduction,retail staff can quickly master the use of the system,and the compliance rate of cigarette display continues to improve,laying the foundation for widespread application.Thus,the CDIGS system provides a new way for cigarette display supervision,effectively guiding retailers to autonomously arrange cigarette displays according to rules and offering a new means for tobacco companies to quickly obtain relevant marketing data from retailers.关键词
卷烟陈列/卷烟识别/警示标志检测/翻拍检测/深度学习Key words
cigarette display/cigarette recognition/warning signs detection/Re-photography detection/deep learning引用本文复制引用
李焕森,覃忠达,谭春,林光耀,秦岑,赵贝贝,黄庆庆..基于深度学习技术的卷烟陈列智能化督导的研究与应用[J].中国烟草学报,2024,30(4):113-122,10.基金项目
中国烟草总公司广西壮族自治区公司科技计划项目"'终端之眼':零售终端卷烟陈列督导智能化研究"(201945000024132) （201945000024132）
