Automatic Detection of Surface Damage and Discoloration in Transmission-line Insulators using YOLOv4
Keywords:
Anchor clustering, CSPDarknet-53, Deep learning, Discoloration classification, Object detection, Power-system condition monitoring, Surface damage detection, Transmission-line insulator, UAV-assisted inspection, YOLOv4Abstract
Surface deterioration of overhead transmission-line insulators poses a well-documented threat to grid reliability, yet timely identification of cracks, broken sheds, and contamination-induced discoloration remains operationally challenging because conventional visual inspection is slow, subjective, and hazardous at tower height. This study describes an end-to-end deep-learning inspection framework built on the YOLOv4 (You Only Look Once version 4) single-stage detector, targeting three semantically distinct classes: intact insulator body, mechanically damaged region, and discolored surface patch. The discoloration class captures an early-stage degradation indicator that precedes structural fracture and is routinely missed by binary defect classifiers. A purpose-assembled dataset of 1,247 field images was gathered from pole-mounted cameras and UAV (Unmanned Aerial Vehicle) platforms, preprocessed with per-channel z-score normalization, and augmented through mosaic composition, random scale jitter, color-space perturbation, and Gaussian noise injection to mitigate class imbalance and viewpoint variability. Anchor clusters were derived by k-means on the annotated bounding-box population to replace the ImageNet-tuned defaults, and training was conducted on CSPDarknet-53 (Cross Stage Partial Darknet-53) with the complete YOLOv4 bag-of-freebies schedule for 300 epochs. On the held-out test partition, the detector achieved an average precision of 96.47% for the insulator class, 99.17% for the damaged-part class, and an overall mean average precision of 97.82% at an IoU (Intersection over Union) threshold of 0.50, while sustaining 43.2 frames per second on the training hardware. A structured comparison against Faster R-CNN (region-based convolutional neural network), SSD (single shot multibox detector), and YOLOv3 baselines confirms a 4.1-percentage-point mAP (mean average precision) gain relative to the nearest single-stage competitor at comparable latency. These findings establish YOLOv4 as a practically viable baseline for automated insulator health monitoring and motivate further work toward edge-device deployment and multi-degradation severity scoring.
References
R. Gorur, E. Cherney, and R. Hackam, “The AC and DC Performance of Polymeric Insulating Materials Under Accelerated Aging in a Fog Chamber,” IEEE Trans. Power Del., vol. 3, no. 4, pp. 1892–1902, 1988.
Z. A. Siddiqui and U. Park, “A drone based transmission line components inspection system with deep learning technique,” Energies, vol. 13, no. 13, p. 3348, 2020.
J. Katrasnik, F. Pernus, and B. Likar, “A Survey of Mobile Robots for Distribution Power Line Inspection,” IEEE Trans. Power Del., vol. 25, no. 1, pp. 485–493, 2010.
J. Wang, C. Liu, T. Fu, and L. Zheng, “Research on automatic target detection and recognition based on deep learning,” Journal of Visual Communication and Image Representation, vol. 60, pp. 44–50, Apr. 2019.
S. Ren, K. He, R. Girshick, and J. Sun, “Faster R-CNN: Towards Real-Time Object Detection With Region Proposal Networks,” IEEE Trans. Pattern Anal. Mach. Intell., vol. 39, no. 6, pp. 1137–1149, 2017.
J. Redmon, S. Divvala, R. Girshick, and A. Farhadi, “You Only Look Once: Unified, Real-Time Object Detection,” in Proc. IEEE Conf. Comput. Vis. Pattern Recognit. (CVPR), Las Vegas, NV, USA, pp. 779–788, 2016.
W. Liu et al., “SSD: Single Shot MultiBox Detector,” in Proc. Eur. Conf. Comput. Vis. (ECCV), pp. 21–37, 2016.
J. Redmon and A. Farhadi, “YOLOv3: An Incremental Improvement,” arXiv preprint arXiv:1804.02767, 2018.
A. Bochkovskiy, C.-Y. Wang, and H.-Y. M. Liao, “YOLOv4: Optimal Speed and Accuracy of Object Detection,” arXiv preprint arXiv:2004.10934, 2020.
Z. Qiu, X. Zhu, C. Liao, D. Shi, and W. Qu, “Detection of Transmission Line Insulator Defects Based on an Improved Lightweight YOLOv4 Model,” Appl. Sci., vol. 12, no. 3, p. 1207, 2022.
Y. Chen, H. Liu, J. Chen, J. Hu, and E. Zheng, “Insu-YOLO: An Insulator Defect Detection Algorithm Based on Multiscale Feature Fusion,” Electronics, vol. 12, no. 15, p. 3210, 2023.
L. Zhang, B. Li, Y. Cui, Y. Lai, and J. Gao, “Research on Improved YOLOv8 Algorithm for Insulator Defect Detection,” J. Real-Time Image Process., vol. 21, art. 22, 2024.
Z. Xu and X. Tang, “Transmission line insulator defect detection algorithm based on MAP-YOLOv8,” Sci. Rep., vol. 15, p. 10288, 2025.
X. Liu, X. Miao, H. Jiang, and J. Chen, “Data analysis in visual power line inspection: An in-depth review of deep learning for component detection and fault diagnosis,” Annu. Rev. Control, vol. 50, pp. 253–277, 2020.
X. Tao, D. Zhang, Z. Wang, X. Liu, H. Zhang, and D. Xu, “Detection of Power Line Insulator Defects Using Aerial Images Analyzed With Convolutional Neural Networks,” IEEE Trans. Syst., Man, Cybern.: Syst., vol. 50, no. 4, pp. 1486–1498, 2020.
Z. Zhang, S. Huang, Y. Li, H. Li, and H. Hao, “Image Detection of Insulator Defects Based on Morphological Processing and Deep Learning,” Energies, vol. 15, no. 7, p. 2465, 2022.
Q. Wen, Z. Luo, R. Chen, Y. Yang, and G. Li, “Deep Learning Approaches on Defect Detection in High Resolution Aerial Images of Insulators,” Sensors, vol. 21, no. 4, p. 1033, 2021.
W. Chen, Y. Li, and Z. Zhao, “Missing-Sheds Granularity Estimation of Glass Insulators Using Deep Neural Networks Based on Optical Imaging,” Sensors, vol. 22, no. 5, p. 1737, 2022.
T. Wang, Y. Zhai, Y. Li, W. Wang, G. Ye, and S. Jin, “Insulator Defect Detection Based on ML-YOLOv5 Algorithm,” Sensors, vol. 24, no. 1, p. 204, 2024.
N. Zhang, J. Su, Y. Zhao, and H. Chen, “Insulator-YOLO: Transmission Line Insulator Risk Identification Based on Optimized YOLOv5,” Processes, vol. 12, no. 11, p. 2552, 2024.
H. Ru, W. Zhang, G. Wang et al., “OMAL-YOLOv8: Real-time detection algorithm for insulator defects based on optimized feature fusion,” J. Real-Time Image Process., vol. 22, p. 52, 2025.
Z. Jing, “Insulator detection study based on improved Faster-RCNN,” in 3D Imaging—Multidimensional Signal Processing and Deep Learning, vol. 1, pp. 141–152, Jul. 2022.
C. Lanhang, “Defect Identification of Electricity Transmission Line Insulators Based on an Improved YOLOv4 Model,” Heliyon, vol. 10, no. 10, e30405, 2024.
S. Dian, X. Zhong, and Y. Zhong, “Faster R-Transformer: An efficient method for insulator detection in complex aerial environments,” Measurement, vol. 199, p. 111238, Aug. 2022.
X. Miao, X. Liu, J. Chen, S. Zhuang, J. Fan, and H. Jiang, “Insulator Detection in Aerial Images for Transmission Line Inspection Using Single Shot Multibox Detector,” IEEE Access, vol. 7, pp. 9945–9956, 2019.
