改进的MobileNetV3热轧钢带表面缺陷分类算法  

Improved MobileNetV3 hot-rolled strip surface defect classification algorithm

作　　者：熊政 车文刚[1] 保永莉 刘晓彤 XIONG Zheng;CHE Wengang;BAO Yongli;LIU Xiaotong(Faculty of Information Engineering and Automation,Kunming University of Science and Technology,Kunming 650504,China;School of Life Science and Technology,Kunming University of Science and Technology,Kunming 650504,China)

机构地区：[1]昆明理工大学信息工程与自动化学院,云南昆明650504 [2]昆明理工大学生命科学与技术学院,云南昆明650504

出　　处：《陕西理工大学学报（自然科学版）》2023年第5期30-37,共8页Journal of Shaanxi University of Technology:Natural Science Edition

基　　金：云南省大学生创新创业训练计划项目(2021106740066)。

摘　　要：热轧制品在工业制造领域有着重要作用,针对其在生产中的表面缺陷问题,提出了一个基于MobileNetV3算法改进的热轧钢带表面缺陷分类算法,通过引入改进的置换注意力机制SAMM替换原模型中的SE注意力机制,在降低参数量的同时提高分类准确率。同时,引入Mish激活函数来替换原模型中的Hard-Swish激活函数,进一步提高分类准确率。将两者结合,构造新的Bneck模块,在此基础上构造新的A-MobileNetV3热轧钢带表面缺陷分类算法。使用NEU-CLS热轧钢带表面缺陷数据集进行分类实验,实验结果表明,改进后的算法分类准确率达到了99.44%,相比于原模型提高了5.74%,而参数量仅为10.35 MB,相比于原模型减少了35.59%。表明改进后的分类算法可以胜任热轧钢带表面缺陷分类的任务。Hot rolled products play an important role in the field of industrial manufacturing.To solve the problem of surface defects in production,an improved classification algorithm for surface defects of hot rolled strips based on MobileNetV3 algorithm is proposed.By introducing the improved replacement attention mechanism SAMM to replace the SE attention mechanism in the original model,the classification accuracy is improved while the quantity of parameters is reduced.At the same time,Mish activation function is introduced to replace the Hoard-Swish activation function in the original model to further improve the classification accuracy.Combining the two,a new Bneck module is constructed,on which a new A-MobileNetV3 hot-rolled steel strip surface defect classification algorithm is constructed.Classification experiments using NEU-CLS hot-rolled strip surface defect dataset show that the improved algorithm achieves a classification accuracy of 99.44%,an improvement of 5.74% compared with the original model,and the parameter amount is only 10.35 MB,which is 35.59% less than the original model.The results show that the improved classification algorithm can be used to classify surface defects of hot-rolled steel strips.

关 键 词：图像分类 MobileNetV3算法 注意力机制 激活函数 

分 类 号：TP391.4[自动化与计算机技术—计算机应用技术]