Scholarlink

<p>Person re-identification (Re-ID) plays a vital role in smart security and video surveillance systems. However, it remains a challenging task due to factors such as occlusion, visual similarity among individuals, and cluttered backgrounds. Existing networks often extract relatively simple features, which limits their ability to distinguish pedestrians in complex environments due to insufficient discriminative power. To address these challenges, we propose an efficient Feature Association Attention Network (FAA-Net) that integrates both local and global features. Specifically, we design a local-global feature association (LGFA) attention mechanism, which combines spatial and channel domain attention in a complementary manner to enhance the extraction of discriminative features. By effectively associating local details with global context, FAA-Net captures key visual cues and transforms them into more distinctive feature representations. Extensive experiments are conducted on four standard benchmark datasets: Market-1501, DukeMTMC-ReID, CUHK-03, and MSMT17. The results demonstrate that FAA-Net consistently outperforms state-of-the-art methods, especially in challenging Re-ID scenarios. In addition, comprehensive ablation studies validate the effectiveness of each component in our proposed architecture.</p>

错误：搜索内容不能为空，请输入英文关键词
错误：关键词超出字数限制，请精简
高级检索

FAA-Net: enhancing person re-identification through local-global feature association attention

  • 期刊论文

  • 发表日期：

    2025年9月15日

查看全文
  • Yangqi Zheng,
  • Liang Zhang,
  • Jun Liang

摘要

Person re-identification (Re-ID) plays a vital role in smart security and video surveillance systems. However, it remains a challenging task due to factors such as occlusion, visual similarity among individuals, and cluttered backgrounds. Existing networks often extract relatively simple features, which limits their ability to distinguish pedestrians in complex environments due to insufficient discriminative power. To address these challenges, we propose an efficient Feature Association Attention Network (FAA-Net) that integrates both local and global features. Specifically, we design a local-global feature association (LGFA) attention mechanism, which combines spatial and channel domain attention in a complementary manner to enhance the extraction of discriminative features. By effectively associating local details with global context, FAA-Net captures key visual cues and transforms them into more distinctive feature representations. Extensive experiments are conducted on four standard benchmark datasets: Market-1501, DukeMTMC-ReID, CUHK-03, and MSMT17. The results demonstrate that FAA-Net consistently outperforms state-of-the-art methods, especially in challenging Re-ID scenarios. In addition, comprehensive ablation studies validate the effectiveness of each component in our proposed architecture.