Scholarlink

<p>In response to the challenges of detecting flames and safety helmets in complex substation environments, this paper proposes a novel deep learning model, YOLO-DySE. Substations present unique challenges due to their intricate layouts, variable lighting conditions (day, night, or low-light), and dynamic environmental factors such as seasonal changes and equipment aging, which complicate the accurate detection of critical hazards, including flames, and the need for safety helmets. These hazards are critical because undetected flames can lead to catastrophic fires, resulting in significant economic losses and safety risks. At the same time, improper helmet usage increases the risk of personnel injury in high-voltage environments. The proposed YOLO-DySE model introduces a dynamic adaptive data augmentation technique (DADE) that adjusts training data in real time, significantly surpassing traditional data augmentation methods in adaptability and generalization in such complex settings. In the feature extraction stage, the model incorporates the C3-SE module, which employs a dynamic weighting mechanism to enhance feature extraction and improve the accuracy of target localization. Additionally, the DyLAMHead module utilizes multiple lightweight attention mechanisms to dynamically adjust attention weights and an inter-layer interaction module to enhance information fusion, improving multi-scale feature processing and overall performance. Experimental results demonstrate that YOLO-DySE significantly outperforms the baseline model, achieving a recall rate of 0.876, with a notable 9.9% improvement in flame detection recall. These advancements make YOLO-DySE highly suitable for ensuring safety in critical substation infrastructure.</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

Multi-target detection for safety monitoring in complex substation environments using YOLO-DySE

  • 期刊论文

  • 发表日期：

    2025年6月26日

查看全文
  • Jiang Junjie,
  • Zhang Yongqi,
  • Wan Anping,
  • Khalil AL-Bukhaiti,
  • Junhao Huang,
  • Xiaomin Cheng

摘要

In response to the challenges of detecting flames and safety helmets in complex substation environments, this paper proposes a novel deep learning model, YOLO-DySE. Substations present unique challenges due to their intricate layouts, variable lighting conditions (day, night, or low-light), and dynamic environmental factors such as seasonal changes and equipment aging, which complicate the accurate detection of critical hazards, including flames, and the need for safety helmets. These hazards are critical because undetected flames can lead to catastrophic fires, resulting in significant economic losses and safety risks. At the same time, improper helmet usage increases the risk of personnel injury in high-voltage environments. The proposed YOLO-DySE model introduces a dynamic adaptive data augmentation technique (DADE) that adjusts training data in real time, significantly surpassing traditional data augmentation methods in adaptability and generalization in such complex settings. In the feature extraction stage, the model incorporates the C3-SE module, which employs a dynamic weighting mechanism to enhance feature extraction and improve the accuracy of target localization. Additionally, the DyLAMHead module utilizes multiple lightweight attention mechanisms to dynamically adjust attention weights and an inter-layer interaction module to enhance information fusion, improving multi-scale feature processing and overall performance. Experimental results demonstrate that YOLO-DySE significantly outperforms the baseline model, achieving a recall rate of 0.876, with a notable 9.9% improvement in flame detection recall. These advancements make YOLO-DySE highly suitable for ensuring safety in critical substation infrastructure.