A method and system for target recognition in SAR images

By fusing sub-aperture decomposition and multi-branch coordinate attention mechanism into SAR images, strong scattering points and multi-scale feature maps are extracted, solving the problem of high computational complexity in SAR image target recognition and improving recognition capability and physical interpretability.

CN122244663APending Publication Date: 2026-06-19NAT UNIV OF DEFENSE TECH
0 Cites 0 Cited by

Patent Information

Authority / Receiving Office
CN · China
Patent Type
Applications(China)
Current Assignee / Owner
NAT UNIV OF DEFENSE TECH
Filing Date
2025-07-28
Publication Date
2026-06-19

AI Technical Summary

⚠Technical Problem

Existing SAR image target recognition methods have high computational complexity in feature extraction and are difficult to combine physical interpretability with parametric representation capabilities.

⚗Method used

By performing sub-aperture decomposition on SAR images to obtain multi-directional sub-aperture images, strong scattering point feature maps and multi-scale feature maps in the image domain are extracted. Then, a multi-branch coordinate attention mechanism is used for fusion and feature extraction, which reduces computational complexity and improves the model's recognition ability and physical interpretability.

🎯Benefits of technology

This approach achieves improved model recognition capability and physical interpretability of SAR image target identification while reducing computational complexity, and enhances the accuracy and robustness of feature extraction.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure 1
    Figure 1
  • Figure 2
    Figure 2
Patent Text Reader

Abstract

This application discloses a SAR image target recognition method and system. The method includes: obtaining multi-directional SAR sub-aperture images by performing sub-aperture decomposition on SAR images; obtaining strong scattering point feature maps based on the SAR sub-aperture images; obtaining image domain multi-scale feature maps based on the SAR images; fusing and extracting features from the strong scattering point feature maps and the image domain multi-scale feature maps using a multi-branch coordinate attention mechanism to obtain target recognition regions; by using the strong scattering point feature maps as salient features in scattering information to correspond to the distribution of complex local structures of the target, it can characterize both geometric structure information and reflect the scattering differences of different targets. Introducing it into a deep network can simultaneously improve the model's recognition ability and physical interpretability, while reducing the computational complexity of SAR image feature extraction; the system also has the same beneficial effects.
Need to check novelty before this filing date? Find Prior Art