基于改进的Bi-LSTM的三维模型检索方法研究  

Research on 3D model retrieval method based on improved Bi⁃LSTM

作　　者：马腾 Ma Teng(College of Information Engineering,North China University of Water Resources and Electric Power,Zhengzhou 450000,China)

机构地区：[1]华北水利水电大学信息工程学院,郑州450000

出　　处：《现代计算机》2023年第23期27-31,共5页Modern Computer

摘　　要：设计了一种基于改进的双向长短时记忆网络的三维模型检索方法。首先,利用卷积神经网络提取三维模型视图的低级特征;然后,使用改进的双向长短时记忆网络详细地学习视图之间的关系,提高提取的特征的质量;最后,将得到的特征经过全连接层生成三维模型描述符用于检索。检索时通过计算两个模型之间的欧氏距离来计算相似度。该方法的三维模型分类准确率为96.9%,检索mAP为86.8%,实现了优异的性能。This paper designed a 3D model retrieval method based on improved bidirectional long short-term memory network.Firstly,a convolutional neural network is used to extract the low-level features of the 3D model views,and then a bidirectional long short-term memory network with coordinate attention mechanism is used to learn the relationship between views in detail to improve the quality of the extracted features.Finally,the obtained features were passed through the fully connected layer to gener-ate 3D model descriptors for retrieval.The similarity is calculated by calculating the Euclidean distance between the two models during retrieval.The 3D model classification accuracy of the proposed method is 96.9%,and the retrieval mAP is 86.8%,achieving excellent performance.

关 键 词：三维模型检索 多视图 深度学习 双向长短时记忆 坐标注意力 

分 类 号：TP391.41[自动化与计算机技术—计算机应用技术] TP183[自动化与计算机技术—计算机科学与技术]