基于改进GhostNet的联排塑瓶微孔检测算法  

Microporous detection algorithm of row plastic bottles based on improved GhostNet

作　　者：刘澎笠 任德均[1] 曹林杰 农皓程 黄德成 LIU Pengli;REN Dejun;CAO Linjie;NONG Haocheng;HUANG Decheng(School of Mechanical Engineering,Sichuan University,Chengdu 610065,China)

机构地区：[1]四川大学机械工程学院,四川成都610065

出　　处：《传感器与微系统》2023年第5期103-107,共5页Transducer and Microsystem Technologies

摘　　要：联排塑瓶包装完整性检测领域,常用高压放电法进行微孔缺陷检测。针对现有的阈值比较法几乎无法检出2μm以下微孔缺陷的问题,分析了基于介质阻挡放电(DBD)的高压放电法电气学模型,发现了检测波形在微孔区域会上下振荡的特征。提出一种基于改进GhostNet的高压放电微孔检测算法。使用Sym2小波基对原始电流数据进行分解重构提取该关键特征,再通过短时傅里叶变换(STFT)生成二维时频图谱作为网络输入。在GhostNet的原型上将G-bneck裁剪到5个,将第1个G-bneck的卷积核尺寸改为1×1,第5个改为5×5。结果表明:优化后的算法测试集准确率达到99.1%,召回率高达99.7%,具有较好的检测效果。In the field of packaging integrity testing of tandem plastic bottles,microporous defect detection is commonly carried out by high-voltage discharge method.Aiming at the problem that the existing threshold comparison method can hardly detect microporous defects below 2μm,the electrical model of high-voltage discharge method based on dielectric barrier discharge(DBD)is analyzed,and the characteristics of the detection waveform oscillating up and down in the micropore region are found.A high-voltage discharge microhole detection algorithm based on improved GhostNet is proposed.The key feature is extracted by decomposition and reconstruction of the original current data by using the sym2 wavelet base,and then a two-dimensional time-frequency map is generated by short-time Fourier transform(STFT)as the network input.Crop the G-bneck to 5 on the prototype of GhostNet,change the convolution kernel size of the first G-bneck to 1×1,and change of the fifth to 5×5.The results show that the accuracy of the optimized algorithm test set reaches 99.1%,and the recall rate is as high as 99.7%,which has a good detection effect.

关 键 词：包装完整性测试 高压检漏 小波变换 GhostNet模块 介质阻挡放电 

分 类 号：TP399[自动化与计算机技术—计算机应用技术]