Wind Turbine Blade Damage Detection Based on the Improved YOLOv5 Algorithm
Yuying Zhang, Long Wang, Chao Huang, Xiong Luo · 2023
Regular inspection and maintenance of wind turbine blades can effectively avoid possible structural failures of wind turbines. A large number of high-resolution images of wind turbines can be obtained through drone inspection shots. This experiment performs data pre-processing and manual annotation of wind turbine blade damage for these images, and is based on YOLOv5 for the object detection of the blade damage. The experimental results show that the model can eventually predict the location and class of blade damage with almost human-level accuracy. It is further shown that for the smaller training set like this experiment, image enhancement before training can better improve the prediction accuracy.