基于Transformer的SAR图像飞机检测识别
Aircraft Detection and Recognition in SAR Images Based on Transformer
邓鑫 1向聪 1张俊 1王伟1
作者信息
- 1. 西安电子工程研究所 西安 710100
- 折叠
摘要
目前深度学习技术已经广泛应用于合成孔径雷达(SAR)目标检测和识别相关任务中,但是由于SAR图像成像噪点较多,物体边缘轮廓不够清楚,对检测造成了一定的困难.鉴于此,提出了一种基于Transformer的SAR飞机目标检测模型.首先,根据SAR图像的特征采用Swin-Transform-er作为基准网络,该网络不仅可以增强网络的全局特征提取能力,且相比于Vison-Transformer有更低的复杂度.其次,为增强模型的边缘特征提取能力,提出了一种边缘检测模块.最后,为增强模型的多尺度特征提取能力,同时保证模型不会出现过拟合问题,提出了 GFPN 结构来实现特征融合.
Abstract
Deep learning technology has been widely used in synthetic aperture radar(SAR)target detection and recognition tasks.However,due to the high noise level and unclear object edge contours in SAR images,target de-tection is challenging.In this paper,a SAR aircraft target detection model based on transformer is proposed.First,considering the characteristics of SAR images,Swin-Transformer is used as the benchmark network,which not only enhances the network's global feature extraction capability but also has lower complexity compared to Vision-Trans-former.Second,an edge detection module is proposed to enhance the edge feature extraction capability of the mod-el.Finally,a GFPN structure is proposed for feature fusion to enhance the model's multi-scale feature extraction ca-pability while ensuring that the model does not suffer from overfitting issues.
关键词
Transformer/特征融合/SAR/目标检测Key words
Transformer/feature fusion/SAR/target detection引用本文复制引用
出版年
2024
NETL
NSTL
万方数据