微型电脑应用2026，Vol.42Issue(4)：101-104,110,5.
融合注意力机制的非完整档案跨模态数据哈希检索方法
A Cross-modal Data Hash Retrieval Method for Incomplete Archives Incorporating Attention Mechanism
摘要
Abstract
Incomplete archives often exhibit high complexity and incompleteness due to data loss,diverse formats,and informa-tion fragmentation between modals,which makes difficult to conduct efficient and accurate cross-modal retrieval directly.To narrow the semantic gap between modals,a cross-modal data hash retrieval method for incomplete archives incorporating atten-tion mechanism is proposed.This paper combines deep neural networks to extract features from images and text separately,treats image features as queries and text features as keys and values,and combines multi-head attention mechanism for fusion processing.By calculating the distance between different feature elements and the upper and lower boundaries of the dimension,and combining the comparison operation,the vector elements are hashed and encoded.By calculating the Hamming distance be-tween the query vector and the archive vector,the search results are sorted based on similarity scores.The retrieval accuracy of the proposed method is tested,and the test results show that the P-R curve of the method performs well and has ideal retrieval performance when using the proposed method for hash retrieval of cross-modal data.关键词
非完整档案/跨模态检索/注意力机制/哈希检索/CNN模型Key words
incomplete archives/cross-modal retrieval/attention mechanism/hash retrieval/CNN model分类
信息技术与安全科学引用本文复制引用
王督,杜嘉程,高伟,孟凡平..融合注意力机制的非完整档案跨模态数据哈希检索方法[J].微型电脑应用,2026,42(4):101-104,110,5.基金项目
国网河南省电力公司科技项目(112377836) （112377836）
