-
摘要:
绝缘子作为电力系统安全运行的核心部件，其缺陷的及时发现对电网可靠性至关重要，而无人机巡检作为低空电力监测的主流方式，对搭载的检测模型提出了轻量化与高精度的双重要求。以YOLOv11为基础模型，提出一种基于无人机视角的轻量级绝缘子缺陷检测算法。在YOLOv11骨干网络中，融合MobileNetV4通用倒瓶颈结构的改进特征提取单元，增强对绝缘子细微缺陷的感知能力；在YOLOv11颈部网络引入了层级化空间筛选特征金字塔网络，通过优化跨层特征交互路径降低模型参数冗余；在检测输出端，采用动态可变形卷积检测头替代传统检测模块，提升对缺陷几何形变的自适应能力。实验结果表明：所提轻量级模型与YOLO系列模型相比，在保证检测精度的基础上，参数量能够降低12.35%以上，更适用于无人机、巡检机器人等边缘设备，为输电线路绝缘子缺陷的实时检测提供了高效解决方案。
Abstract:As the primary method of low-altitude power monitoring, unmanned aerial vehicle (UAV) inspection presents the dual requirements of lightweight and high precision for the detection model. The timely detection of defects is crucial to the reliability of the power grid since it is the fundamental component of the safe operation of the power system. Based on YOLOv11, a lightweight insulator defect detection algorithm based on the UAV perspective is proposed. Firstly, in the YOLOv11 backbone network, the improved feature extraction unit of MobileNetV4, with a general inverted bottleneck structure, is integrated to enhance the perception of subtle defects of insulators. Secondly, the YOLOv11 neck network was integrated with a hierarchical spatial screening feature pyramid network, and the interaction path of cross-layer features was optimized to minimize model parameter redundancy. Finally, at the detection output, the dynamic deformable convolution detection head is used to replace the traditional detection module to improve the adaptability to the geometric deformation of defects. Experimental results show that compared with the YOLO series model, the proposed lightweight model can reduce the number of parameters by more than 12.35% on the basis of ensuring detection accuracy, which is more suitable for edge equipment such as UAVs and inspection robots, and provides an efficient solution for the real-time detection of transmission line insulator defects.
-
Key words:
- insulator /
- defect detection /
- feature enhancement /
- lightweight /
- YOLOv11
-
图 3 不同检测算法的缺陷识别效果示例
Figure 3. Example diagram of defect recognition effect of different detection algorithms
表 1 平台配置参数
Table 1. Platform configuration parameters
GPU CPU 学习框架 GPU并行计算库 每批次样本数 学习率 优化器算法 迭代次数 NVIDIA GeForce RTX 3070 Intel(R) Core(TM) i7-10700 CPU @ 2.90 GHz PyTorch Cuda11.8 16 0.01 SGD 300 
下载: 导出CSV
表 2 对比实验结果
Table 2. Comparative experimental results
算法 P/% R/% mAP50-95/% 参数量 YOLOv5n 89.7 89.5 50.0 2.51×106 YOLOv6n 87.8 87.1 49.2 4.24×106 YOLOv8n 96.1 89.5 50.9 3.01×106 YOLOv10n 91.8 87.9 50.5 2.71×106 YOLOv11n 91.6 89.2 49.2 2.58×106 IDD-YOLOv5 90.3 85.7 44.6 3.82×106 LiteYOLO 86.8 83.7 42.7 3.74×106 本文 92.1 85.8 50.2 2.20×106 注：加粗数值表示性能最优。
下载: 导出CSV
表 3 消融实验结果
Table 3. Ablation experimental results
UIB HSFPN DynamicDCNv3Head P/% R/% mAP50-95/% 参数量 91.6 89.2 49.2 2.58×106 √ 92.6 85.1 49.3 2.55×106 √ 89.6 85.3 48.2 1.85×106 √ 89.0 88.6 49.6 2.42×106 √ √ 89.2 86.7 46.6 1.86×106 √ √ 90.8 84.5 48.8 2.37×106 √ √ 90.3 87.6 49.9 2.23×106 √ √ √ 92.1 85.8 50.2 2.20×106
下载: 导出CSV
-
[1] Li D H, Lu Y, Gao Q, et al. LiteYOLO-ID: a lightweight object detection network for insulator defect detection[J]. IEEE Transactions on Instrumentation and Measurement, 2024, 73: 5023812. [2] Liu J, Hu M M, Dong J Y, et al. Summary of insulator defect detection based on deep learning[J]. Electric Power Systems Research, 2023, 224: 109688. [3] Liu Y, Liu D C, Huang X B, et al. Insulator defect detection with deep learning: a survey[J]. IET Generation, Transmission & Distribution, 2023, 17(16): 3541-3558. [4] Wang S Q, Liu Y F, Qing Y H, et al. Detection of insulator defects with improved ResNeSt and region proposal network[J]. IEEE Access, 2020, 8: 184841-184850. [5] 韩玉洁, 曹杰, 刘琨, 等. 基于改进YOLO的无人机对地多目标检测[J]. 电子测量技术, 2020, 43(21): 19-24.Han Y J, Cao J, Liu K, et al. UAV ground multi-target detection based on improved YOLO[J]. Electronic Measurement Technology, 2020, 43(21): 19-24(in Chinese). [6] 贾玉进, 张振程, 李浠铭, 等. 基于轻量化YOLOv5网络的输电线路绝缘子缺陷检测[J]. 电力学报, 2024, 39(1): 36-44.Jia Y J, Zhang Z C, Li X M, et al. Defect detection of insulator on transmission line based on lightweight YOLOv5 network[J]. Journal of Electric Power, 2024, 39(1): 36-44(in Chinese). [7] Zhang S H, Qu C N, Ru C Y, et al. Multi-objects recognition and self-explosion defect detection method for insulators based on lightweight GhostNet-YOLOV4 model deployed onboard UAV[J]. IEEE Access, 2023, 11: 39713-39725. [8] 陈佳韵, 肖根福, 张祥明. 基于改进YOLOv7-tiny的自爆绝缘子检测算法[J]. 电子测量技术, 2025, 48(7): 66-74.Chen J Y, Xiao G F, Zhang X M. Detection algorithm for self-exploding insulator based on improved YOLOv7-tiny[J]. Electronic Measurement Technology, 2025, 48(7): 66-74(in Chinese). [9] Chen Y F, Zhang C Y, Chen B, et al. Accurate leukocyte detection based on deformable-DETR and multi-level feature fusion for aiding diagnosis of blood diseases[J]. Computers in biology and medicine, 2024, 170: 107917. [10] Qin D F, Leichner C, Delakis M, et al. MobileNetV4: universal models fortheMobile ecosystem[C]//Proceedings of the Computer Vision-ECCV 2024. Berlin: Springer, 2025: 78-96. [11] Wang W H, Dai J F, Chen Z, et al. InternImage: exploring large-scale vision foundation models with deformable convolutions[C]//Proceedings of the 2023 IEEE/CVF Conference on Computer Vision and Pattern Recognition. Piscataway: IEEE Press, 2023: 14408-14419. [12] 路成龙, 冯月贵, 庆光蔚. 基于改进YOLOv8的电梯内电动车识别方法研究[J]. 机械制造与自动化, 2024, 53(4): 219-223.Lu C L, Feng Y G, Qing G W. Research on identification method of electric vehicles in elevators based on improved YOLOv8[J]. Machine Building & Automation, 2024, 53(4): 219-223(in Chinese). [13] 侯卫民, 何孟玲, 赵梦瑶, 等. 无人机视角下施工现场工人防护用具检测方法研究[J]. 计算机工程与应用, 2025, 61(15): 353-362.Hou W M, He M L, Zhao M Y, et al. Research on detection method of protective equipment for construction site workers from UAV perspective[J]. Computer Engineering and Applications, 2025, 61(15): 353-362(in Chinese). [14] 何飞熊, 谢海巍, 蒲超, 等. 基于改进YOLOv8网络的道路病害检测方法[J]. 计算机与现代化, 2025(2): 108-113.He F X, Xie H W, Pu C, et al. A new method of pavement disease detection based on improved YOLOv8[J]. Computer and Modernization, 2025(2): 108-113(in Chinese). [15] Dai X Y, Chen Y P, Xiao B, et al. Dynamic head: unifying object detection heads with attentions[C]//Proceedings of the 2021 IEEE/CVF Conference on Computer Vision and Pattern Recognition. Piscataway: IEEE Press, 2021: 7369-7378. [16] 刘柏甫. 面向三维点云数据的目标检测方法研究[D]. 北京: 北京交通大学, 2023.Liu B F. Research on object detection method for 3D point cloud data[D]. Beijing: Beijing Jiaotong University, 2023(in Chinese). [17] 孙光灵, 黄磊, 吴倩. 基于KD-YOLO的轻量化安全帽佩戴检测方法[J]. 淮北师范大学学报(自然科学版), 2024, 45(2): 41-48.Sun G L, Huang L, Wu Q. Lightweight safety helmet wearing detection method based on KD-YOLO[J]. Journal of Huaibei Normal University (Natural Science), 2024, 45(2): 41-48(in Chinese). [18] Khanam R, Hussain M. YOLOv11: an overview of the key architectural enhancements[EB/OL]. 2024: arXiv: 2410.17725. https://arxiv.org/abs/2410.17725. [19] Jiang P Y, Ergu D, Liu F Y, et al. A review of YOLO algorithm developments[J]. Procedia Computer Science, 2022, 199: 1066-1073. [20] Hussain M. YOLO-v1 to YOLO-v8, the rise of YOLO and its complementary nature toward digital manufacturing and industrial defect detection[J]. Machines, 2023, 11(7): 677. [21] Lu Y, Li D H, Gao Q, et al. IDD-YOLOv5: a lightweight insulator defect real-time detection algorithm[C]//Proceedings of the 2024 IEEE International Conference on Mechatronics and Automation. Piscataway: IEEE Press, 2024: 491-495. -
下载:
点击查看大图
计量
- 文章访问数: 288
- HTML全文浏览量: 173
- PDF下载量: 8
- 被引次数: 0
