Image-Based Condition Monitoring of Transmission Line Conductors Using Image Processing and Deep Neural Networks

Transmission high voltage lines play a key role in maintaining the link between the generation and distribution sectors in the power system. Since they are typically located in harsh external environments, they are exposed to different natural phenomena which cause conductor degradation in mechanisms such as annealing, corrosion, and fretting fatigue. The inability to detect these failures at their rudimentary stages can eventually lead to a loss of reliability of the power system and safety concerns. However, trending inspection methods based on Unmanned Aerial Vehicles (UAV) can be used for effective routine checks of these conductors if the image processing techniques are properly utilized. Therefore, this paper presents a baseline framework for image-based online condition monitoring using RGB and IR images. The competency of extracting macroscopic features from these images is well investigated in this paper, and a foundation is laid to conduct further research that idealizes a more practical alternative for the condition assessment of conductors. Mainly, this work showcases a conductor defect localization and classification method using thresholding and the YOLOv7 algorithm with the aid of an artificially created defect dataset that mimics real-life cases. Moreover, Convolutional Neural Network (CNN) was applied to perform an age-based classification despite the illuminance level of the moment in which the RGB image was taken. The classifier has been shown to distinguish the three interested classes with an accuracy of 98.5%. In the final stage, thermal image-based analysis was carried out which supports the aforementioned classifier through validation and provides a basis for effectively localizing the corroded areas of a transmission line conductor according to heat dissipation. Further developments in the discussed approaches will advance the power system to a more cost-effective, reliable, and safer state.