基于并联注意力机制的多视图三维模型特征提取  

Multi-view 3D model feature extraction based on parallel attention mechanism

作　　者：王家欣 WANG Jiaxin(College of Information,North China University of Water Resources and Electric Power,Zhengzhou 450018,China)

机构地区：[1]华北水利水电大学信息工程学院,郑州450018

出　　处：《计算机应用文摘》2025年第6期57-59,62,共4页

摘　　要：在当前基于视图的三维模型检索技术领域,大多数方法主要集中于提取视图的全局特征,而较少考虑局部特征信息及其与其他视图之间的关联性。为了解决这一问题,提出了一种新的特征提取技术。该技术基于深度学习框架,应用卷积神经网络,并引入并联注意力机制,以增强特征的区分能力。通过在ModelNet40数据库上进行的实验结果表明,采用该技术对三维模型的多视图进行输入,并在网络层面嵌入并联注意力模块和双向长短时记忆网络进行特征提取和分类时,分类精度超过了现有主流算法。In the current field of view based 3D model retrieval technology,most methods mainly focus on extracting global features of views,and pay less attention to local feature information and its correlation with other views.To address this issue,a new feature extraction technique has been proposed.This technology is based on a deep learning framework,applies convolutional neural networks,and introduces parallel attention mechanisms to enhance the discriminative ability of features.The experimental results conducted on the ModelNet40 database show that when using this technology to input multiple views of 3D models and embedding parallel attention modules and bidirectional long short-term memory networks for feature extraction and classification at the network level,the classification accuracy exceeds that of existing mainstream algorithms.

关 键 词：三维模型检索 特征提取 卷积神经网络 并联注意力机制 双向长短时记忆网络 

分 类 号：TP389[自动化与计算机技术—计算机系统结构]