Scholarlink

<p>In the domain of power transmission, external damage hazards within transmission corridors significantly compromise the secure operation of the power grid. Therefore, it is of great importance to perceive the status and behavior. Given the complex environment surrounding transmission corridors, the types and scales of external damage hazards pose significant detection challenges. Thus, this study combined the improved YOLOv5s detection model with the lightweight StrongSORT tracking model for real-time perception of external damages. In the target detection phase, we construct the YOLO-CS-ASFF model based on the YOLOv5s architecture which uses the ConNeXt module with the fused SimAM attention mechanism to extract crucial features in complex backgrounds. The ASFF module enhances damage perception across different scales by optimizing feature fusion networks. Additionally, we use the SIoU loss function to improve the precision of external damage detection. In the target tracking phase, we optimize the appearance branching network of StrongSORT using the full-scale network (OSNet), which enhances the capability of StrongSORT to meet the real-time inspection requirements. Experimental results show that the improved YOLO-CS-ASFF achieved a mean Average Precision (mAP) and Recall of 92.8% and 85.3%, respectively, with an improvement of 3.24% and 1.32%. The StrongSORT tracking model attained tracking accuracy and precision of 63.3% and 78.9%, respectively, with a detection speed increase of 5.5 frames per second. The model effectively addresses the ID switching problem of obscured hazard targets, and improve the robustness of external breakage hazard tracking. The proposed method provides a technical reference for real-time perception of external breakage hazards in actual transmission corridors.</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

Research on perceptual methods of external damage hazards for transmission corridors

  • 期刊论文

  • 发表日期：

    2025年1月17日

查看全文
  • He Su,
  • Jiaomin Liu,
  • Zhenzhou Wang,
  • Pingping Yu,
  • Yuting Yan

摘要

In the domain of power transmission, external damage hazards within transmission corridors significantly compromise the secure operation of the power grid. Therefore, it is of great importance to perceive the status and behavior. Given the complex environment surrounding transmission corridors, the types and scales of external damage hazards pose significant detection challenges. Thus, this study combined the improved YOLOv5s detection model with the lightweight StrongSORT tracking model for real-time perception of external damages. In the target detection phase, we construct the YOLO-CS-ASFF model based on the YOLOv5s architecture which uses the ConNeXt module with the fused SimAM attention mechanism to extract crucial features in complex backgrounds. The ASFF module enhances damage perception across different scales by optimizing feature fusion networks. Additionally, we use the SIoU loss function to improve the precision of external damage detection. In the target tracking phase, we optimize the appearance branching network of StrongSORT using the full-scale network (OSNet), which enhances the capability of StrongSORT to meet the real-time inspection requirements. Experimental results show that the improved YOLO-CS-ASFF achieved a mean Average Precision (mAP) and Recall of 92.8% and 85.3%, respectively, with an improvement of 3.24% and 1.32%. The StrongSORT tracking model attained tracking accuracy and precision of 63.3% and 78.9%, respectively, with a detection speed increase of 5.5 frames per second. The model effectively addresses the ID switching problem of obscured hazard targets, and improve the robustness of external breakage hazard tracking. The proposed method provides a technical reference for real-time perception of external breakage hazards in actual transmission corridors.