ACFM: Adaptive Channel Feature Matching for Pedestrian Re-Identification
Image misalignment is a significant challenge in the field of pedestrian re-identification. Previous studies typically align pedestrian features using additional models or by leveraging auxiliary information. However, these methods are data-dependent and can fail when dealing with substantial scene...
Saved in:
| Main Authors: | , |
|---|---|
| Format: | Article |
| Language: | English |
| Published: |
IEEE
2025-01-01
|
| Series: | IEEE Access |
| Subjects: | |
| Online Access: | https://ieeexplore.ieee.org/document/10994765/ |
| Tags: |
Add Tag
No Tags, Be the first to tag this record!
|
| _version_ | 1850136772476928000 |
|---|---|
| author | Zhengcai Lu Zhengwei Tian |
| author_facet | Zhengcai Lu Zhengwei Tian |
| author_sort | Zhengcai Lu |
| collection | DOAJ |
| description | Image misalignment is a significant challenge in the field of pedestrian re-identification. Previous studies typically align pedestrian features using additional models or by leveraging auxiliary information. However, these methods are data-dependent and can fail when dealing with substantial scene variations. To address these challenges, this paper proposes an efficient adaptive channel feature matching (ACFM) strategy. This method adaptively aligns channel feature maps according to image content, enables accurate matching of local features across images, and mitigates the need for supplementary data. Building on the ACFM approach, this paper develops a multi-branch adaptive matching network model. The model incorporates a channel attention mechanism to enhance the representation capacity of individual channels and integrates the ACFM algorithm within local branches to optimize the distance metric computation, effectively capturing and aligning local details. The multi-branch structure is designed to handle both global and local features, enabling the network to comprehensively capture and integrate image information. Evaluation experiments conducted on two widely-used benchmark datasets, Market-1501 and DukeMTMC-ReID, demonstrate that the proposed method offers significant advantages over current state-of-the-art approaches. |
| format | Article |
| id | doaj-art-e1bc474ac69f4338a1268f0c3c30400f |
| institution | OA Journals |
| issn | 2169-3536 |
| language | English |
| publishDate | 2025-01-01 |
| publisher | IEEE |
| record_format | Article |
| series | IEEE Access |
| spelling | doaj-art-e1bc474ac69f4338a1268f0c3c30400f2025-08-20T02:31:02ZengIEEEIEEE Access2169-35362025-01-0113822788229010.1109/ACCESS.2025.356880610994765ACFM: Adaptive Channel Feature Matching for Pedestrian Re-IdentificationZhengcai Lu0https://orcid.org/0009-0002-4408-7243Zhengwei Tian1School of Artificial Intelligence and Big Data, Luzhou Vocational and Technical College, Luzhou, Sichuan, ChinaSchool of Artificial Intelligence and Big Data, Luzhou Vocational and Technical College, Luzhou, Sichuan, ChinaImage misalignment is a significant challenge in the field of pedestrian re-identification. Previous studies typically align pedestrian features using additional models or by leveraging auxiliary information. However, these methods are data-dependent and can fail when dealing with substantial scene variations. To address these challenges, this paper proposes an efficient adaptive channel feature matching (ACFM) strategy. This method adaptively aligns channel feature maps according to image content, enables accurate matching of local features across images, and mitigates the need for supplementary data. Building on the ACFM approach, this paper develops a multi-branch adaptive matching network model. The model incorporates a channel attention mechanism to enhance the representation capacity of individual channels and integrates the ACFM algorithm within local branches to optimize the distance metric computation, effectively capturing and aligning local details. The multi-branch structure is designed to handle both global and local features, enabling the network to comprehensively capture and integrate image information. Evaluation experiments conducted on two widely-used benchmark datasets, Market-1501 and DukeMTMC-ReID, demonstrate that the proposed method offers significant advantages over current state-of-the-art approaches.https://ieeexplore.ieee.org/document/10994765/Pedestrian re-identificationadaptive channel feature matchingmulti-branch structurechannel attention mechanism |
| spellingShingle | Zhengcai Lu Zhengwei Tian ACFM: Adaptive Channel Feature Matching for Pedestrian Re-Identification IEEE Access Pedestrian re-identification adaptive channel feature matching multi-branch structure channel attention mechanism |
| title | ACFM: Adaptive Channel Feature Matching for Pedestrian Re-Identification |
| title_full | ACFM: Adaptive Channel Feature Matching for Pedestrian Re-Identification |
| title_fullStr | ACFM: Adaptive Channel Feature Matching for Pedestrian Re-Identification |
| title_full_unstemmed | ACFM: Adaptive Channel Feature Matching for Pedestrian Re-Identification |
| title_short | ACFM: Adaptive Channel Feature Matching for Pedestrian Re-Identification |
| title_sort | acfm adaptive channel feature matching for pedestrian re identification |
| topic | Pedestrian re-identification adaptive channel feature matching multi-branch structure channel attention mechanism |
| url | https://ieeexplore.ieee.org/document/10994765/ |
| work_keys_str_mv | AT zhengcailu acfmadaptivechannelfeaturematchingforpedestrianreidentification AT zhengweitian acfmadaptivechannelfeaturematchingforpedestrianreidentification |