基于多特征融合神经网络的遥感影像推荐方法  

Remote Sensing Image Recommendation Method Based on Multi-feature Fusion Neural Network

作　　者：王梅瑞 楚博策 孔二旦 陈金勇[1] 朱进 李峰[1] 王士成[1] WANG Meirui;CHU Boce;KONG Erdan;CHEN Jinyong;ZHU Jin;LI Feng;WANG Shicheng(The 54th Research Institute of CETC,Shijiazhuang 050081,China)

机构地区：[1]中国电子科技集团公司第五十四研究所,河北石家庄050081

出　　处：《无线电工程》2024年第2期378-389,共12页Radio Engineering

基　　金：河北省省级科技计划(21340302D)。

摘　　要：针对目前遥感用户获取遥感影像时采用的人工查询和订购方式较为低效的问题,提出了一种基于多特征融合神经网络的遥感影像个性化推荐框架。设计遥感影像语义体系实现影像语义提取,结合用户操作记录与职责描述文本构建遥感领域知识图谱。通过嵌入表征方法提取用户与影像的多维度属性特征。设计了一种基于协同过滤的多特征融合神经网络模型,借助神经网络的高维度空间建模能力实现用户与影像多维属性特征的有效融合,达到较好的配对推荐效果。实验结果显示,相较于先前的推荐方法,所提方法的性能获得显著提高,有效提升了遥感数据服务的精准度与时效性。Considering the problem that manual query and ordering methods used by remote sensing users to obtain remote sensing images are relatively inefficient,a personalized recommendation framework for remote sensing images based on multi-feature fusion neural network is proposed.Firstly,this method designs a remote sensing image semantic system to realize image semantic extraction,and combines user operation records and responsibility description texts to construct a remote sensing knowledge graph.Subsequently,multi-dimensional attribute characteristics of users and images are extracted by embedding characterization methods.Finally,a multi-feature fusion neural network based on collaborative filtering is designed.With the help of neural network s high-dimensional spatial modeling capability,the proposed model achieves effective fusion of multi-dimensional attribute features of users and images,and obtains a better matching recommendation effect.Experimental results show the significant improvement of the proposed method compared with previous recommendation methods,which effectively improves the accuracy and timeliness of remote sensing services.

关 键 词：主动推荐 遥感影像 特征融合 神经网络 

分 类 号：TP391.3[自动化与计算机技术—计算机应用技术]