Transmission Line Vibration Damper Detection Using Deep Neural Networks Based on UAV Remote Sensing Image

Vibration dampers can greatly eliminate the galloping phenomenon of overhead transmission wires caused by wind. The detection of vibration dampers based on visual technology is an important issue. The current vibration damper detection work is mainly carried out manually. In view of the above situation, this article proposes a vibration damper detection model named DamperYOLO based on the one-stage framework in object detection. DamperYOLO first uses a Canny operator to smooth the overexposed points of the input image and extract edge features, then selectees ResNet101 as the backbone of the framework to improve the detection speed, and finally injects edge features into backbone through an attention mechanism. At the same time, an FPN-based feature fusion network is used to provide feature maps of multiple resolutions. In addition, we built a vibration damper detection dataset named DamperDetSet based on UAV cruise images. Multiple sets of experiments on self-built DamperDetSet dataset prove that our model reaches state-of-the-art level in terms of accuracy and test speed and meets the standard of real-time output of high-accuracy test results.

Automated summary

This article proposes a vibration damper detection model named DamperYOLO based on visual technology, which improves detection speed and accuracy through edge features and attention mechanism. Multiple sets of experiments on a dataset built from UAV images prove that the model reaches the state-of-the-art level and meets the standard of real-time output of high-accuracy test results.