Resnet18模型在玻璃表面脏污识别中的应用  

Application of Resnet18 in the Identification of Glass Surface Dirt

作　　者：苏国兴 常光超 任传成[1] SU Guoxing;CHANG Guangchao;REN Chuancheng(School of Computer and Information Science,Dezhou University,Dezhou,China,253023;School of Information and Control Engineering,Jilin University of Chemical Technology,Jilin,China,132022)

机构地区：[1]德州学院计算机与信息学院,山东德州253023 [2]吉林化工学院信息与控制工程学院,吉林吉林132022

出　　处：《福建电脑》2024年第9期17-21,共5页Journal of Fujian Computer

基　　金：德州学院大学生创新创业训练校级项目“基于图像识别技术的玻璃脏污识别系统研究”(No.X202310448061)资助。

摘　　要：为解决高层建筑窗户的玻璃表面赃污识别检测的困难,本文提出一种基于Resnet18的玻璃表面脏污识别方法。该方法以玻璃表面脏污层为研究对象,将迁移学习和计算机视觉算法相结合,实现对玻璃表面赃污的识别和检测。测试的结果表明,Resnet18在自建玻璃表面脏污数据集上的识别准确率达到了99.10%,可以有效地对玻璃表面脏污进行识别,对实际应用具有一定的借鉴意义。To solve the difficulty of identifying and detecting glass surface dirt on high-rise building windows,this paper proposes a glass surface dirt identification method based on Resnet18.Taking the dirty layer on the surface of glass as the research object,transfer learning and computer vision algorithms are combined to achieve recognition and detection of dirt on the glass surface.The test results show that Resnet18 has a recognition accuracy of 99.10%on the self built glass surface dirt dataset,which can effectively identify glass surface dirt and has certain reference significance for practical applications.

关 键 词：玻璃脏污层 图像识别 残差网络模型 迁移学习 

分 类 号：TP391[自动化与计算机技术—计算机应用技术]