Catenary insulator surface defect detection is of great significance to ensure the safe operation of the traction power supply system. The insulator images collected by the railway 4C inspection vehicle have problems such as weak edge features, significant scale differences, and high environmental noise, and the accuracy and stability of the intelligent detection method still need improvement. Therefore, an adaptive multi-scale calibration fusion(AMCF) YOLO algorithm for surface defect detection of a cantilever insulator is proposed. The algorithm adopts targeted feature enhancement, feature fusion, and regression optimization measures, and the problem of insufficient feature fusion between model levels is improved by designing a hierarchical feature coordination pyramid network. The P5 layer feature integrated with the deformable attention mechanism is used for module extraction, and combined with lightweight down-sampling with edge-enhanced branches, the number of algorithm parameters is reduced, while key defect information is retained. The enhanced loss function is introduced to strengthen the regression constraint on the medium quality prediction box and improve the positioning accuracy of the model. Comparative experiments show that the AMCF-YOLO parameter is only 4.2 M, the mAP50 reaches 93.4%, and the mAP50-95 is 60.1%, which are 3.3 and 5.6 percentage points higher than the benchmark model YOLOv11n, respectively. The generalization ability experiments on three public datasets verify the effectiveness and stability of the proposed model. The developed insulator surface multi-defect detection system is deployed on edge devices. The test results show that the detection accuracy of the proposed model is stable, and the detection frame rate reaches 65 frame/s, which meets the requirements of real-time detection on-site.
式中:g(a, b)=max(0,1-|a-b|); “:”表示索引(rx, ry )遍历特征图 z ∈RH×W×C 的所有位置. 由于g仅在(px,py )最近的四个整数坐标处非零, 所以可简化为这四个位置上的加权平均. 对查询 q 、键 k 、值 v 实施多头注意力机制m, 并引入相对位置偏移 R . 每个注意力头的输出表示为:
BIANJ P, LIF, HAOP X, et al .Damage identification and location of transmission line insulator in complex environment[J].High Voltage Engineering, 2022, 48(2):681-688.(in Chinese)
ZHANGT R, DUANC C, HUY W, et al .A YOLOv7-OpenCV algorithm for detecting the size of insulator damage in overhead contact system[J].Journal of Railway Science and Engineering, 2025, 22(4):1870-1882.(in Chinese)
ABULIZIJ, LIUY L, CAOL, et al. Study of insulator fault detection algorithm based on improved Cascade R-CNN network[J].Journal of Electric Power Science and Technology, 2023, 38(3): 140-148. (in Chinese)
ZHANGF Z, SONGL, LIR H, et al .Evaluation method of defect and fracture reason for composite insulator[J]. High Voltage Engineering, 2012, 38(11):3093-3100.(in Chinese)
HUX J, NIC, WUC P, et al. Lightweight insulator defect detection algorithm based on UAV perspective[J/OL]. Journal of Beijing University of Aeronautics and Astronautics, 1-7.(2025-09-10)[2025-09-25].in Chinese)
WANGY Q, YUANT, NIEL, et al .Machine vision inspection method for defects of glass insulator[J]. High Voltage Engineering, 2022, 48(12): 4933-4940.(in Chinese)
ZHUE W, LIANGZ, XIAOJ W, et al .Research of small object detection problem based on MDS-YOLO model[J]. Journal of Hunan University (Natural Sciences), 2024, 51(12):78-86.(in Chinese)
ZHANGG N, LIUZ G, HANY, et al. A fast fuzzy matching method of fault detection for rod insulators of high-speed railways[J].Journal of the China Railway Society, 2013, 35(5):27-33.(in Chinese)
YANGH M, LIUZ G, HANY, et al .Defective condition detection of insulators in electrified railway based on feature matching of speeded-up robust features[J].Power System Technology, 2013, 37(8):2297-2302.(in Chinese)
[19]
WUX, YUANP, PENGQ, et al. Detection of bird nests in overhead catenary system images for high-speed rail[J]. Pattern Recognition, 2016, 51: 242-254.
LIUC Y, WUY Q, LIUJ J. Research progress of deep learning methods for insulator defect detection in UAV based aerial images[J]. Transactions of China Electrotechnical Society, 2025, 40(9): 2897-2916.(in Chinese)
LIP, SUY L, NINGH, et al. Self-explosion defect detection of power insulators based on embedded YOLO network[J]. Journal of Electrical Technology, 2025,40(23):7806-7818.(in Chinese)
[24]
TANP, LIX F, DINGJ . et al. Mask R-CNN and multifeature clustering model for catenary insulator recognition and defect detection[J]. Journal of Zhejiang University-Science A, 2022, 23 (9): 745-757.
[25]
WENQ D, LUOZ Q, CHENR T, et al. Deep learning approaches on defect detection in high resolution aerial images of insulators[J]. Sensors, 2021, 21(4): 1033.
LIX, LIUS N, YANGZ, et al. Multi-target detection of transmission lines based on improved Cascade R-CNN[J]. Journal of Electronic Measurement and Instrumentation, 2021, 35(10): 24-32. (in Chinese)
TANGB N, LIUM X, LIH, et al. Insulator defect detection algorithm based on improved YOLO v10[J/OL]. China Southern Power Grid Technology, 1-11.(2025-03-25)[2025-12-17]. in Chinese)
SONGZ W, HUANGX B, JIC, et al. Insulator defect detection and fault warning method for transmission line based on flexible YOLOv7[J]. High Voltage Technology, 2023, 49(12): 5084-5094. (in Chinese)
LIB, QUL Y, ZHUX S, et al .Insulator defect detection based on multi-scale feature fusion[J]. Transactions of China Electrotechnical Society, 2023, 38(1): 60-70.(in Chinese)
YANGX, XIONGW, MENGS Z, et al. Lightweight insulator defect detection algorithm based on YOLOv8 [J/OL]. Advances in laser and optoelectronics, 1-17.(2025-05-08)[2025-12-17]. in Chinese)
JIANGX J, WANGR T, MAY H .Insulator defect detection based on lightweight improved RT-DETR edge deployment algorithm[J].Transactions of China Electrotechnical Society, 2025, 40(3): 842-854.(in Chinese)
LUOX L, WANGY B, PUY Y, et al .Road disease detection RGT-YOLOv7 model under multiple diseases complicated scenarios[J].Journal of Hunan University (Natural Sciences), 2024, 51(12): 107-118.(in Chinese)
[40]
XIAZ F, PANX R, SONGS J, et al .Vision transformer with deformable attention[EB/OL]. (2022-01-03)[2025-12-17].
[41]
LIUC, WANGK G, LIQ, et al .Powerful-IoU:more straightforward and faster bounding box regression loss with a nonmonotonic focusing mechanism[J]. Neural Networks, 2024, 170:276-284.
[42]
TAOX, ZHANGD P, WANGZ H, et al. Detection of power line insulator defects using aerial images analyzed with convolutional neural networks[J]. IEEE Transactions on Systems, Man, and Cybernetics:Systems, 2020, 50(4): 1486-1498.
RENM, ZHANGX F, ZHIX B, et al .An annotated street view image dataset for automated road damage detection[J]. Scientific Data, 2024, 11:407.
[45]
WANGC C, HEW, NIEY, et al .Gold-YOLO:efficient object detector via gather-and-distribute mechanism[EB/OL]. (2023-10-23)[2025-12-17].
[46]
TANM X, PANGR M, LEQ V .EfficientDet:scalable and efficient object detection[C]//2020 IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR), June 13-19, 2020.Seattle, WA, USA.IEEE, 2020:10778-10787.
[47]
OUYANGD L, HES, ZHANGG Z, et al. Efficient multi-scale attention module with cross-spatial learning[EB/OL]. (2023-06-06)[2025-12-17].
[48]
XUW, WANY, ZHAOD. SFA:efficient attention mechanism for superior CNN performance[J]. Neural Processing Letters, 2025, 57(2):38.
[49]
WUZ Y, DINGT J, LUY F, et al .Token statistics transformer:linear-time attention via variational rate reduction[EB/OL]. (2024-12-23)[2025-12-23].
[50]
ZHANGX, SONGY Z, SONGT T, et al. LDConv:linear deformable convolution for improving convolutional neural networks[J].Image and Vision Computing, 2024, 149:105190.
[51]
SUNKARAR, LUOT. No more strided convolutions orpooling: a new CNN building block for low-resolution images and small objects[C]//Joint European Conference on Machine Learning and Knowledge Discovery in Databases. Cham:Springer Nature Switzerland, 2022: 443-459.
[52]
WILLIAMST, LIR Y. Wavelet pooling for convolutional neural networks[C]//International Conference on Learning Representations, 2018.
基金资助
国家自然科学基金资助项目(51567014)
National Natural ScienceFoundation of China(51567014)
甘肃省自然科学基金资助项目(24JRRA1175)
Natural Science Foundation of Gansu Province(24JRRA1175)
甘肃省青年科技基金资助项目(20JR5RA382)
Scientific Foundation for Young Scholars of Gansu Province(20JR5RA382)
甘肃省计量研究院科技计划资助项目(GJY2024002)
Science and Technology Foundation of Measurement Institute of Gansu Province(GJY2024002)