Scholarlink

<p>This paper presents a multimodal behavior recognition method that fuses infrared and visible images to address image degradation problems caused by poor lighting and floating dust in mining environments. The proposed Low-Light Enhancement Fusion Network (LIEFusion) integrates a Low-Light Enhancement Network (LIENet) and a Texture Contrast Enhancement Network (TCENet) to enhance degraded visible-light images and produce high-quality fused results. Furthermore, a YOLOE-based human behavior recognition model (YOLOE-HSB) is developed, which combines YOLOE detection with spatiotemporal skeleton features for improved abnormal behavior recognition. Experimental results demonstrate that the proposed method achieves an average accuracy improvement of <InlineEquation ID="IEq1"> <EquationSource Format="TEX">\(4.3\%\)</EquationSource> <EquationSource Format="MATHML"><math> <mrow> <mn>4.3</mn> <mo>%</mo> </mrow> </math></EquationSource> </InlineEquation> on public datasets and <InlineEquation ID="IEq2"> <EquationSource Format="TEX">\(10.2\%\)</EquationSource> <EquationSource Format="MATHML"><math> <mrow> <mn>10.2</mn> <mo>%</mo> </mrow> </math></EquationSource> </InlineEquation> on a self-built dual-modal underground abnormal behavior dataset, significantly improving recognition accuracy and reducing false alarms compared with single-modal approaches.</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

Underground personnel abnormal behavior recognition method based on infrared and visible image fusion

  • 期刊论文

  • 发表日期：

    2026年1月22日

查看全文
  • Qiang Guo,
  • Junming Bai,
  • Hongguang Pan,
  • Huipeng Zhang,
  • Ze Jiang,
  • Libin Zhang,
  • Li Li

摘要

This paper presents a multimodal behavior recognition method that fuses infrared and visible images to address image degradation problems caused by poor lighting and floating dust in mining environments. The proposed Low-Light Enhancement Fusion Network (LIEFusion) integrates a Low-Light Enhancement Network (LIENet) and a Texture Contrast Enhancement Network (TCENet) to enhance degraded visible-light images and produce high-quality fused results. Furthermore, a YOLOE-based human behavior recognition model (YOLOE-HSB) is developed, which combines YOLOE detection with spatiotemporal skeleton features for improved abnormal behavior recognition. Experimental results demonstrate that the proposed method achieves an average accuracy improvement of \(4.3\%\) 4.3 % on public datasets and \(10.2\%\) 10.2 % on a self-built dual-modal underground abnormal behavior dataset, significantly improving recognition accuracy and reducing false alarms compared with single-modal approaches.