融合多维度特征的绝缘子状态边缘识别方法

针对传统的绝缘子状态识别方法存在实时性差、特征提取能力不足的问题，基于边缘计算的思想，提出了一种融合多维度特征的绝缘子状态边缘识别方法。利用云边协同和边边联邦协同的联合技术手段，构建了绝缘子状态的边缘识别框架。设计了一种融合多维度特征提取的深度学习网络，该网络采用ResNet101作为主干特征提取网络，使用Inception模块构建数据池化层，嵌入压缩激励模块和卷积注意力模块，从不同维度对特征进行高效提取。采用包括正常和缺陷2种状态的数据集进行绝缘子状态边缘识别实验，平均识别准确率达到了99%。实验表明了融合多维度特征的绝缘子状态边缘识别方法的有效性。

An Edge Recognition Method for Insulator State Based on Multi-dimension Feature Fusion

Traditional insulator state recognition methods have such problems as poor real-time performance and insufficient feature extraction ability. Based on the idea of edge computing, this paper proposes a method for recognizing the insulator state based on multi-dimension feature fusion. An edge recognition framework for insulator state is constructed using cloud edge collaboration and edge federation collaboration. And a deep learning network integrating multi-dimension feature extraction is designed, which, by using the ResNet101 as the main feature extraction network, uses the Inception module to build the data pooling layer, and embeds the compression incentive module and convolution attention module to extract features from different dimensions. An insulator state recognition experiment is conducted using the data set of normal and defect states, and the average recognition accuracy reaches 99%. The experimental results have proved the validity of the proposed method.