多模态数据共享子空间的锌析出工况分类方法  

A Classification Method for Zinc Precipitation Conditions Based on Multimodal Data Shared Subspace

作　　者：陈莉莉 CHEN Li-li(Hunan Industry Polytechnic,Changsha 410208,China)

机构地区：[1]湖南工业职业技术学院,湖南长沙410208

出　　处：《电脑与电信》2024年第12期43-48,共6页Computer & Telecommunication

基　　金：湖南省自然科学基金,项目编号:2022JJ60029。

摘　　要：为了解决锌电解极板析出工况分类中红外图像和日志数据特征维度不一致导致的融合建模难问题,提出了一种基于多模态数据共享子空间方法——ZOA-ShareSpace。该方法以日志文本和红外图像为输入,通过潜在共享编码器提取出不同模态信息的潜在共享信息,并利用共享激活函数将潜在共享信息映射到共享空间中。实验结果表明,该方法具有较高的准确率,能够有效地实现对锌析出工况的分类,为生产实践提供了有力支持。In order to solve the problem of fusion modeling caused by the inconsistency of the feature dimensions of infrared images and log data in the classification of precipitation conditions of zinc electrolytic plates,a multi-modal data sharing subspace method based on ZOA-ShareSpace is proposed.This method uses log text and infrared images as inputs and extracts latent shared information from different modalities through a latent shared encoder.A shared activation function is then used to map the latent shared information into a shared space.Experimental results show that this method achieves high accuracy and effectively classifies zinc deposition conditions,providing strong support for production practices.

关 键 词：ZOA-ShareSpace 多模态共享子空间 工况分类 锌电解 

分 类 号：TP311.13[自动化与计算机技术—计算机软件与理论]