[1]BAI Jianpeng,WANG Wei,CHEN Yuxi,et al.Detection and spatial location of wind turbine blades based on lightweight YOLOv5[J].CAAI Transactions on Intelligent Systems,2022,17(6):1173-1181.[doi:10.11992/tis.202204016]
Copy
CAAI Transactions on Intelligent Systems[ISSN 1673-4785/CN 23-1538/TP] Volume:
17
Number of periods:
2022 6
Page number:
1173-1181
Column:
学术论文—智能系统
Public date:
2022-11-05
- Title:
-
Detection and spatial location of wind turbine blades based on lightweight YOLOv5
- Author(s):
-
BAI Jianpeng; WANG Wei; CHEN Yuxi; JIAO Songming
-
Department of Automation, North China Electric Power University, Baoding 071003, China
-
- Keywords:
-
wind turbine; unmanned aerial vehicle; object detection; YOLOv5; lightweight; deep learning; blade tip; accurate positioning
- CLC:
-
TP138;TP242
- DOI:
-
10.11992/tis.202204016
- Abstract:
-
The application of unmanned aerial vehicles (UAVs) for autonomous inspection of wind turbines requires precise positioning of the paddle blade tips, but the detection efficiency of conventional target detection algorithms is low due to the limited computing power of the onboard computer. Therefore, a method of the blade and spatial location detection of wind turbines based on lightweight YOLOv5 is proposed. Initially, the YOLOv5 target detection algorithm is lightly improved using ShuffleNetv2 as the feature extraction backbone network. The algorithm is then used to detect the hub and blades of the turbine in the panoramic image to obtain the pixel coordinates of the hub and blade tips. Finally, the UAV positional information and geometric relationship between the spatial planes are used to accurately locate the wind turbine blades. The tests show that the improved target detection algorithm with 1.536 × 106 parameters on the DJI MANIFOLD2-C improves detection speed by 47%, up to 29.4 f/s. The designed positioning method can accurately locate the tips of wind turbine blades with both horizontal and height positioning errors of ±5 cm and a three-dimensional overall positioning error of ±10 cm.