Image Detection Technology on Pin Defect to Overhead Power Lines
Weidong Zhang, Chaoyue Gu, Zhe Li, Gehao Sheng · 2020 5th Asia Conference on Power and Electrical Engineering (ACPEE) · 2020
To improve efficiency of UAV in routing inspection of overhead power lines as well as detection rate of pin defect to overhead power lines, the thesis proposes a method for detection of pin defect for UAV routing inspection of overhead power lines based on Faster-RCNN algorithm. In view of the fact that UAV routing inspection is characterized by large image background and tiny size of pin, the deep residual network ResNet101 is selected as pre-feature extraction network on the basis of Faster-RCNN and training image scale is increased. Experiments show that the method has a good performance in pin defect detection in UAV patrol image on test set. Compared with other prevailing object detection methods, the detection effect is better and the generalization ability is stronger.