基于改进YOLO v5的烟包切层断面异物检测方法  

Foreign matter detection method of cigarette packet cutting section using improved YOLO v5

作　　者：沈飞翔 陈成军 王金磊 李东年 代成刚 Shen Feixiang;Chen Chengjun;Wang Jinlei;Li Dongnian;Dai Chenggang(School of Mechanical and Automotive Engineering,Qingdao University of Technology,Qingdao Shandong,266520)

机构地区：[1]青岛理工大学机械与汽车工程学院,山东青岛266520

出　　处：《电子测试》2022年第22期55-58,共4页Electronic Test

基　　金：国家自然科学基金资助(52175471)。

摘　　要：当前主流目标检测网络应用于烟包切层断面异物检测时存在召回率低、小目标异物大量漏检误检的问题,针对这一问题,本研究提出一种烟包切层断面异物检测网络YOLO v5-MFF。提出了多特征提取网络、多阶段并行融合机制,增强了网络对小目标异物的特征提取能力;引入ACON类激活函数,使网络根据数据动态选择激活与否及激活函数的表达形式,增强了网络的特征表达能力。实验表明,本研究提出的YOLO v5-MFF对烟包切层断面异物检测的召回率达到了94.2%,比YOLO v5地提升了了4.6%,明显提升了小目标异物的检测能力。At present, the mainstream target detection network is applied to the detection of foreign matters in cigarette packet cutting section, which has the problems of low recall rate and a large number of missed and false detection of small target foreign matters. Aiming at this problem, this study proposes a cigarette packet cutting section foreign matter detection network YOLO v5-MFF. A multi feature extraction network and multi-stage parallel fusion mechanism are proposed to enhance the feature extraction ability of the network for small target foreign bodies;ACON activation function is introduced to enable the network to dynamically select the activation or not and the expression form of the activation function according to the data, so as to enhance the feature expression ability of the network.The experimental results show that the recall rate of YOLO v5-MFF proposed in this study for the detection of foreign matters in the cut section of cigarette packet is 94.2%, which is 4.6% higher than that of YOLO v5, and significantly improves the detection ability of small target foreign matters.

关 键 词：异物检测 多特征提取网络、多阶段并行融合机制 ACON激活函数 

分 类 号：TH16[机械工程—机械制造及自动化]