郑州大学学报（理学版）2024，Vol.56Issue(2)：73-79,7.DOI:10.13705/j.issn.1671-6841.2022271
CNS-Net:一种循环多注意力特征聚合架构
CNS-Net:a Cyclic Multi-attention Feature Aggregation Architecture
摘要
Abstract
The method based on convolutional neural network(CNN)has advantages in extracting global feature information and local feature information with classification of high-resolution remote sensing ima-ges,but it can not effectively distinguish key information from interference information.An end-to-end CNS-Net network was proposed to extract salient features of images.Firstly,a global enhancement mod-ule(GEM)was designed to show the interdependence between modeling channels,so that the network could selectively extract key areas.Secondly,a multi-staged circular attention module(MCAM)was pro-posed to capture the long-term dependency and context-aware information of feature information.Finally,experiments on four public data sets showed that our method achieved the best classification performance.关键词
卷积神经网络/长短期记忆网络/注意力机制/遥感图像Key words
convolutional neural network/long-short term memory/attention mechanism/remote sens-ing image分类
信息技术与安全科学引用本文复制引用
陈俊松,易积政,陈爱斌..CNS-Net:一种循环多注意力特征聚合架构[J].郑州大学学报（理学版）,2024,56(2):73-79,7.基金项目
湖南省自然科学基金项目(2022JJ31022) （2022JJ31022）
湖南省本科教育改革项目(NJG-2021-0532). （NJG-2021-0532）
