Accurate detection, localization, and classification of defects in power transmission line insulators captured by Unmanned Aerial Vehicles (UAVs) are essential for maintaining the safety and reliability of high-voltage power grids. Traditional methods in this domain have encountered challenges in achieving robust performance, particularly in complex environmental conditions, and have primarily focused on detecting a single type of defect. To overcome these challenges, this paper presents a novel deep learning framework that combines the strengths of three state-of-the-art object detection models: YOLOv8, EfficientDet, and Faster R-CNN. The proposed approach enhances both the localization and classification of multiple insulator defects in aerial imagery. By leveraging the unique capabilities of each model, the framework improves detection accuracy, reduces false positives, and enables effective classification of various defect types, contributing to more efficient and automated maintenance strategies for power transmission infrastructure. The experimental results demonstrate the effectiveness of the integrated framework in detecting and categorizing defects, outperforming existing single-model approaches.

Keywords: Power transmission line, insulators, UAV, defect localization, defect classification, YOLOv8, EfficientDet, Faster R-CNN, deep learning, aerial images, object detection.