科技创新与应用2025，Vol.15Issue(28)：14-18,5.DOI:10.19981/j.CN23-1581/G3.2025.28.003
基于EFE-LinkNet网络的SAR图像道路检测算法
赵冰怡 1方佳宾 1张帅豪 1潘志刚2
作者信息
- 1. 中国科学院空天信息创新研究院,北京 100190||中国科学院大学 电子电气与通信工程学院,北京 100049
- 2. 中国科学院空天信息创新研究院,北京 100190
- 折叠
摘要
Abstract
Traditional SAR(Synthetic Aperture Radar)image road detection algorithms have problems such as being greatly affected by speckle noise,difficulty in utilizing high-frequency image information,and low segmentation accuracy.To solve the above problems,this study improves the Linknet network and designs a road detection method for SAR images that combines efficiency and accuracy:EFE-LinkNet(Efficient Feature Enhancement-LinkNet).First,a local feature and directional information enhancement module is designed to enable the network to capture detailed features and directional information in the image;secondly,a feature fusion module is designed to reduce information loss by efficiently fusing deep features and detailed features;finally,for the sample imbalance problem,a mixed loss function is designed to optimize the model training process.Road detection experiments are carried out using real airborne high-resolution SAR image data.The results show that the proposed algorithm has good detection results and verifies the effectiveness of the algorithm.关键词
合成孔径雷达/道路检测/条带池化/残差单元/轻量化Key words
synthetic aperture radar/road detection/strip pooling/residual element/lightweight分类
信息技术与安全科学引用本文复制引用
赵冰怡,方佳宾,张帅豪,潘志刚..基于EFE-LinkNet网络的SAR图像道路检测算法[J].科技创新与应用,2025,15(28):14-18,5.
