An air-ground pedestrian re-identification method based on frequency domain structure alignment and view angle identity decoupling

By using frequency domain structure alignment and viewpoint identity decoupling, the problem of feature instability caused by viewpoint differences in pedestrian re-identification in open and ground environments is solved, improving the accuracy and feature robustness of cross-viewpoint matching.

CN122135409APending Publication Date: 2026-06-02NANJING UNIV OF INFORMATION SCI & TECH
View PDF 0 Cites 0 Cited by

Patent Information

Authority / Receiving Office
CN · China
Patent Type
Applications(China)
Current Assignee / Owner
NANJING UNIV OF INFORMATION SCI & TECH
Filing Date
2026-05-06
Publication Date
2026-06-02

AI Technical Summary

Technical Problem

Existing pedestrian re-identification methods in open areas struggle to effectively address issues such as geometric deformation, semantic misalignment, and damage to identity information when faced with significant differences in perspective, resulting in low cross-perspective matching accuracy and unstable features.

Method used

By using frequency domain structure alignment and viewpoint identity decoupling, local feature resampling and frequency adaptive routing are first performed in the frequency domain to restore the local semantic topology, separate the identity-biased and viewpoint-biased frequency components, and perform identity reliability fidelity preservation and viewpoint bias recovery to improve cross-viewpoint matching accuracy.

Benefits of technology

It achieves more stable cross-view matching, reduces the loss of key identity information, and improves the feature robustness and accuracy of pedestrian re-identification in both open and ground environments.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN122135409A_ABST
    Figure CN122135409A_ABST
Patent Text Reader

Abstract

This invention discloses a method for re-identifying pedestrians in the air and ground based on frequency domain structure alignment and viewpoint identity decoupling, comprising: acquiring an image of the pedestrian to be identified; extracting a spatial domain feature map using a visual backbone network; performing frequency domain-ready structure alignment on the spatial domain feature map; performing frequency domain transformation on the structure-aligned spatial domain feature map to obtain a frequency domain feature map; constructing a unified frequency semantic representation; performing frequency adaptive routing based on the frequency semantic representation, decomposing the frequency domain features into a partial identity frequency component and a partial viewpoint frequency component; performing identity reliability fidelity processing on the partial identity frequency component; performing viewpoint bias recovery processing on the partial viewpoint frequency component; fusing the fidelity-enhanced partial identity frequency component and the shape identity component in the frequency domain to obtain frequency domain identity features, and extracting spatial domain identity features; fusing the frequency domain identity features and the spatial domain identity features to obtain the final identity; this invention improves accuracy and feature robustness.
Need to check novelty before this filing date? Find Prior Art