基于卷积神经网络的绝缘子RTV喷涂缺陷检测方法

Insulator RTV Spraying Defect Detection Method Based on Convolutional Neural Network

绝缘子室温硫化硅橡胶(room temperature vulcanized silicone rubber,RTV)涂层的二次喷涂可以有效延长绝缘子的使用寿命,保障电力系统的安全运行。针对喷涂质量的评估问题,提出一种基于卷积神经网络的RTV喷涂缺陷自动检测方法:首先构建绝缘子RTV喷涂区域提取网络模型,对绝缘子RTV喷涂区域进行精确分割;然后构建绝缘子RTV喷涂缺陷检测网络模型,对5种喷涂缺陷进行语义分割。实验结果证明,所构建的喷涂缺陷检测网络模型在评价指标上优于主流的语义分割网络,具有良好的性能,能够满足应用需求。

The secondary spraying of room temperature vulcanized silicone rubber(RTV)coating on the insulators can effectively prolong the service life of insulators and ensure safe operation of the power system.Aiming at the evaluation of coating quality,an automatic detection method of RTV coating defects based on convolution neural network(CNN)is proposed.Firstly,the extraction network model of insulator RTV coating area is constructed to accurately segment the insulator RTV coating area.Then,the insulator RTV spray defect detection network model is constructed,and the semantic segmentation of five spray defects is carried out.The experimental results show that the constructed spray defect detection network model is superior to the mainstream semantic segmentation network in terms of evaluation indicators,has good performance and can meet the application requirements.