A Detection Method of Self Explosion Defect of Transmission Line Insulator Based on Cascade R-CNN
Shen Houming, Zhen Wei, Peng Fan, Zhao Chun, Xie Tao, Dong Qin, Xiong Jiajun, Liu Jinjuan · 2022
In this paper, a method for detecting the self explosion defect of transmission line insulator based on Cascade RCNN is proposed. This method is proposed to improve the accuracy of insulator defect detection based on deep learning. The ResNeXt101 network is used as the feature extraction network of this method. This method improves the generalization ability of the network. It reduces the number of super parameters without increasing the network parameters, and reduces the difficulty of network design and computational overhead. It improves the detection accuracy of the self explosion defect of transmission line insulator. This paper selects three common methods for comparison: Cascade RCNN detection method with ResNet101 as feature extraction network, Faster RCNN detection method with ResNeXt101 as feature extraction network and RetineNet detection method with ResNeXt101 as feature extraction network. Through experimental analysis, the proposed method has higher reliability, average accuracy and stronger generalization ability. It significantly improves the detection accuracy of self explosion defects of transmission line insulators.