TSDD: A Two-Step Defects Detection Algorithm in Power Line Inspection

Imagery-based inspection, with its small size, affordability, high-resolution imaging capability, and mobility, has become a preferred method in power line inspections. Meanwhile, artificial intelligence (AI)-enabled defects detection has been extensively studied to further improve the efficiency and accuracy of inspection tasks. However, traditional defect detection methods exhibit a high rate of false detections when applied to real-world utility data, impeding their widespread adoption. In this paper, we propose a two-step defects detection (TSDD) algorithm to mitigate these issues. During the first step, power components are located and classified, followed by removing the background. In the subsequent second step, defects within the power component region are detected. Our experiments demonstrate significant improvements in detecting small defects like broken bells and woodpecker holes, enhancing the accuracy and efficiency of power line inspections.