High-Voltage Power Transmission Tower Detection Based on Faster R-CNN and YOLO-V3

摘要

The power transmission mainly depends on overhead transmission infrastructures, such as towers and lines. Automatic inspection by robots or UAVs for the power transmission infrastructures is an essential way to ensure the safety of power transmission. Automatic detection and classification of the power towers is the prerequisite for automatic inspection. This paper compares two state-of-art deep learning methods to realize the high-voltage power transmission tower detection. We build the dedicated dataset of the power towers for multi-object detection, including data collection, preprocessing and annotation. After that, the models of YOLO-v3 and Faster R-CNN are used to solve multi-object detection on our dataset. The performances of the two models are evaluated under different indicators. It is verified that Faster R-CNN has a better detection performance in accuracy. However, the detection speed of YOLO-V3 model is faster and can be used in real-time detection.