Insulator Defect Recognition Based on Faster R-CNN
Yifan Wang, Zhongxu Li, Xuecheng Yang, Ning Luo, Yu Zhao, Gang Zhou
Abstract
Insulators are important parts to ensure the normal operation of transmission lines. The traditional method is to judge the defect of the insulators through human eyes, which is not only low in efficiency, but also strong and dangerous in work. In this paper, the most representative Resnet-50 and Faster R-CNN frameworks in target classification algorithms are used for training to detect and recognize insulator targets, and introduce target network model detection performance indicators. The results show that this method can judge the defect of insulators well, and has strong robustness and practicability. Reduce manpower and improve accuracy and timeliness.
Topics & Concepts
Robustness (evolution)Computer scienceInsulator (electricity)Artificial intelligenceElectric power transmissionMachine learningPattern recognition (psychology)EngineeringElectrical engineeringChemistryBiochemistryGenePower Line Inspection RobotsAdvanced Neural Network ApplicationsHigh voltage insulation and dielectric phenomena