采用改进YoloV4算法的连接件识别方法  

Connection Recognition Method Using Improved YoloV4 Algorithm

作　　者：李翠明[1] 王龙 徐龙儿 王华[1] LI Cuiming;WANG Long;XU Longer;WANG Hua(School of Mechanical and Electrical Engineering,Lanzhou University of Technology,Lanzhou 730050,China)

机构地区：[1]兰州理工大学机电工程学院,兰州730050

出　　处：《机械科学与技术》2024年第12期2138-2146,共9页Mechanical Science and Technology for Aerospace Engineering

基　　金：国家自然科学基金项目(52265065);甘肃省自然科学基金项目(18JR3RA139)。

摘　　要：为实现连接件的自动化装配与分拣,提出了一种改进的YoloV4算法用于连接件的识别。首先在YoloV4的基础上,将YoloV4中的主干网络CSP-Darknet53替换为轻量级的GhostNet网络,同时把YoloV4中用到的普通卷积替换成深度可分离卷积来进一步减少参数量,并通过K-means++聚类算法来避免K-means聚类算法中的缺点,生成先验框尺寸。试验结果表明,改进后的YoloV4算法的平均精度值高达100%,识别速度得到大幅提高,参数量较YoloV4减少了82%,可提高在嵌入式设备的应用范围,为智能制造提供了技术支持。In order to realize the automatic assembly and sorting of connectors,this paper proposes an improved YoloV4 algorithm for connector identification.First,CSP-Darknet53,the backbone network in YoloV4is replacedby a lightweight GhostNet network.At the same time,the ordinary convolution used in YoloV4is also replaced with a deeply separable convolution to further reduce the number of parameters,and K-means++clustering algorithm is used to avoid the shortcomings of K-means clustering algorithm and generate a priori box size.The experimental results show that the average accuracy of the improved YoloV4 algorithm is as high as 100%,the recognition speed is greatly improved,and the number of parameters is reduced by 82%compared with YoloV4,which can improve the application range of embedded devices and provide technical support for intelligent manufacturing.

关 键 词：目标识别 工业零件 YoLoV4 GhostNet 

分 类 号：TP273.5[自动化与计算机技术—检测技术与自动化装置]