基于深度学习模型与稀疏表示的绝缘子状态分类

Insulator Status Classification Based on Deep Learning Model and Sparse Representation

在输电线路中绝缘子的状态直接影响整个输电系统的可靠性,然而复杂背景和不同光照条件下对于绝缘子的状态检测十分困难。如今计算机视觉辅助方法已被广泛应用于电力系统中。提出一种基于深度学习模型与稀疏表示进行绝缘子状态分类的方法,对于待检测的绝缘子图像,通过Faster-RCNN定位后,采用深度残差网络(ResNet)提取图像特征,最后利用稀疏表示进行绝缘子状态分类。该方法与传统方法相比,对于绝缘子的状态分类具有更高的准确率,准确率可达98.67%。

The status of the insulators in power line can directly affect the reliability of power transmission system.However,the complex background and different light conditions make it difficult to detect insulators.Nowadays,computer vision-aided methods have been widely used in electric,power system.In this paper,we propose a novel insulator status classification approach based on deep learning model and sparse representation to classify the status of insulators.In order to detect the insulator image,the Deep Residual Networks（ResNet）is used to extract features after localizing the position of the insulators with FasterRCNN,then insulator status classification is done with sparse representation.Compared with the traditional methods,our method has higher accuracy for the classification of insulator status and achieves the precision rate of 98.67%.