Research on Infrared Image Extraction and Defect Analysis of Composite Insulator Based on U-net

To improve the efficiency and accuracy of defect analysis for infrared images of composite insulators, the processing flow based on U-net model is proposed. By constructing and training the U-net model, the extraction of composite insulators in infrared images is realized. On this basis, the locating of target insulator is also carried out. Meanwhile, the temperature matrix is derived based on the conversion of infrared image and the temperature value mapping is performed to obtain the corresponding temperature value of composite insulator. Then, combined with historical defect characteristics, the defect diagnosis is carried out to automatically determine the type of insulator defect according to the temperature value and the range of temperature rising area. At last, two types of insulator defects are intelligently identified to prove the effectiveness of the proposed method.