Scholarlink

<p>The reliable operation of insulators is essential for maintaining the stability of power systems. Long-term exposure to harsh environmental conditions increases their susceptibility to small-scale faults, including damage and flashover. Timely and accurate fault detection is therefore critical to ensuring reliable power transmission. To address the limitations of existing approaches, particularly image degradation under complex weather conditions such as rain, snow, and fog, as well as insufficient detection accuracy for small targets, an enhanced multi-type insulator fault detection algorithm, termed CIA-YOLO, is proposed. The proposed method is developed based on the YOLOv8 framework and introduces CPA-Enhancer as an adaptive image-processing module to alleviate feature degradation caused by adverse imaging conditions. In the backbone network, a multi-scale feature fusion structure, IMSD-Block, is employed to replace the original C2f Bottleneck module. By leveraging hierarchical feature fusion and a heterogeneous convolution kernel selection mechanism, the proposed structure improves the detection of small-scale and non-structural defects. Within the neck structure, the AR-FPN module integrates re-parameterisation, depth-separable convolutions, and bidirectional feature fusion strategies, thereby enhancing multi-scale feature representation and improving detection efficiency. To comprehensively evaluate performance, a Multi-Weather Insulator Fault Detection Dataset (MTIF-CWE) was constructed to cover diverse weather scenarios. Experimental results indicate that CIA-YOLO increases mAP from 88.1 to 94.5% with only an additional 0.521 million parameters, demonstrating superior capability in detecting small-object faults under complex weather conditions. The proposed approach exhibits strong practical applicability and engineering relevance.</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

CIA-YOLO: multi-type insulator defect detection in transmission systems under sophisticated climatic conditions based on improved YOLOv8 algorithm

  • 期刊论文

  • 发表日期：

    2026年2月25日

查看全文
  • Kangkang Zhang,
  • Jianqiao Ma,
  • Qianyue Wang

摘要

The reliable operation of insulators is essential for maintaining the stability of power systems. Long-term exposure to harsh environmental conditions increases their susceptibility to small-scale faults, including damage and flashover. Timely and accurate fault detection is therefore critical to ensuring reliable power transmission. To address the limitations of existing approaches, particularly image degradation under complex weather conditions such as rain, snow, and fog, as well as insufficient detection accuracy for small targets, an enhanced multi-type insulator fault detection algorithm, termed CIA-YOLO, is proposed. The proposed method is developed based on the YOLOv8 framework and introduces CPA-Enhancer as an adaptive image-processing module to alleviate feature degradation caused by adverse imaging conditions. In the backbone network, a multi-scale feature fusion structure, IMSD-Block, is employed to replace the original C2f Bottleneck module. By leveraging hierarchical feature fusion and a heterogeneous convolution kernel selection mechanism, the proposed structure improves the detection of small-scale and non-structural defects. Within the neck structure, the AR-FPN module integrates re-parameterisation, depth-separable convolutions, and bidirectional feature fusion strategies, thereby enhancing multi-scale feature representation and improving detection efficiency. To comprehensively evaluate performance, a Multi-Weather Insulator Fault Detection Dataset (MTIF-CWE) was constructed to cover diverse weather scenarios. Experimental results indicate that CIA-YOLO increases mAP from 88.1 to 94.5% with only an additional 0.521 million parameters, demonstrating superior capability in detecting small-object faults under complex weather conditions. The proposed approach exhibits strong practical applicability and engineering relevance.