输电线路绝缘子自爆缺陷识别方法

绝缘子自爆缺陷识别是实现运行状态监测和故障诊断的重要前提。针对输电线路航拍图像背景复杂的特点,提出一种基于深度学习的绝缘子自爆缺陷识别方法。本文中应用深度学习目标分类算法中最具代表性的AlexNet,VGG16以及FasterR-CNN框架分别进行分类器和检测器的训练,并将分类器和检测器级联组成级联网络,进行绝缘子目标的检测识别。分类器的正确率达到了72%,检测器的正确率达到了59.6%,级联网络的漏检率降为0。实验结果表明本文的方法能够有效地识别绝缘子、自动化性能良好,为下一步绝缘子故障抢修提供了基础。

Recognition Method for Faults of Insulators on Transmission Lines

The recognition of faults of insulators is a significant precondition for operation state monitoring and fault di. agnosis. In view of the complicated background in the aerial images of transmission lines,a recognition method for faults of insulators based on deep learning is proposed. In this paper,classifiers and detectors are trained using AlexNet,VGG16,and Faster R-CNN,which are typical classification algorithms of deep learning. Moreover,a cascade network is formed by the classifier and detector to perform the detection and identification of insulators. The accuracies for classifier and detector reach 72% and 59. 6%,respectively,and the loss detecting rate of the cascade network is re. duced to 0. Experimental results show that the proposed method can recognize insulators effectively and has satisfying automation performance,which provides a basis for insulator repair at the next step.