SAR image target detection method based on global-local attention and multi-scale fusion enhancement
The SAR image target detection method enhanced by global-local attention and multi-scale fusion solves the problems of noise interference and small target detection in SAR images, and achieves high robustness and high accuracy in target detection, especially for the accurate localization of rotating targets.
Patent Information
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- HANGZHOU DIANZI UNIV
- Filing Date
- 2026-04-17
- Publication Date
- 2026-07-03
AI Technical Summary
SAR image target detection suffers from problems such as severe speckle noise interference, low signal-to-noise ratio, small targets being easily obscured by background clutter, and inaccurate positioning of rotating targets, which are difficult to solve effectively with existing technologies.
A SAR image target detection method based on global-local attention and multi-scale fusion enhancement is adopted, including a continuous denoising attention network, a SAR sensing multi-scale difference fusion module, and a weighted IoU loss function. Through global-local feature fusion and multi-scale feature enhancement, noise is suppressed, and the detection capability of small targets and the positioning accuracy of rotating targets are improved.
It significantly improves the robustness and accuracy of target detection in SAR images, reduces noise interference, enhances the detection capability of small-scale targets and targets with complex backgrounds, and improves the localization accuracy of rotating targets and the convergence speed of the model.
Smart Images

Figure CN122336581A_ABST
Abstract
Citation Information
Patent Citations
Remote sensing image ultra-small target detection method based on improved YOLO
CN121837953A