基于改进YOLOV5的超市商品包装老化自动识别技术  

Automated recognition technology for aging of supermarket product packaging based on improved YOLOV5

作　　者：周昊[1] 陆忞[1] ZHOU Hao;LU Min(State Grid Nanjing Power Supply Company,Nanjing 210019,China)

机构地区：[1]国网南京供电公司,江苏南京210019

出　　处：《粘接》2025年第1期78-81,共4页Adhesion

摘　　要：针对商品包装袋老化图像识别过程中存在的问题,提出基于改进YOLOV5的商品包装袋老化图像识别技术,引入FasterNet轻量级网络,形成C3-Faster模块,进一步减少冗余计算和内存访问,并缩小模型规模。实验结果表明,随着照射时间的延长,吸收峰逐渐增加,老化后官能团的含量逐渐增加。且随着迭代次数大于200时,漏检率呈下降趋势。Yolov5模型的漏检率在迭代800次后趋于收敛,改进Yolov5模型的漏检率在迭代600次后趋于收敛,平均漏检率约为2.66%,远低于Yolov5模型的漏检率。无人超市商品包装袋的破损检测时间及检测准确率效果较优,检测时间为25 ms,较包装袋龟裂及褪色分别降低30.55%及51.92%。In order to solve the problems existing in the process of commodity bag aging image recognition,an im⁃proved commodity bag aging image recognition technology based on YOLOV5 was proposed,and the FasterNet lightweight network was introduced to form the C3-Faster module,which further reduced the number of redundant computation and memory access,and reduced the model size.The experimental results showed that the absorption peaks gradually increased with the prolongation of irradiation time,and the content of functional groups gradually in⁃creased after aging.And when the number of iterations was greater than 200,the missed detection rate showed a downward trend.The leakage rate of Yolov5 model tended to converge after 800 iterations,the leakage rate of the improved Yolov5 model tended to converge after 600 iterations,and the average leakage rate was about 2.66%,which was much lower than that of Yolov5 model.The detection time and detection accuracy of unmanned supermar⁃ket packaging bags were better,with a detection time of 25 ms,which was 30.55%and 51.92%lower than bag cracking and discoloration,respectively.

关 键 词：改进YOLOV5 无人超市 商品包装材料 图像识别 

分 类 号：TQ325.7[化学工程—合成树脂塑料工业] TP391.41[自动化与计算机技术—计算机应用技术]