부산대학교 도서관

Infrared thermal image can record a large number of thermal fault problems in the power grid, and become the most important detection method of thermal defects in current power equipment. At present, the methods of artificial analysis of infrared images and thermal defects of power equipment have some problems, such as low efficiency and error prone, poor real-time performance of massive data input and effective feedback, and difficult data association analysis. In the past, image processing technology was used to automatically analyze the thermal defects of power equipment in the infrared thermal image. Because of the low pixel and fuzzy boundary of the infrared image, the recognition rate of the equipment was low and it was easy to misjudge, which could not meet the actual needs. This paper analyzes and compares the methods used in the past for infrared image recognition of power equipment patrol inspection. On this basis, in order to improve the accuracy of power equipment identification, an infrared image recognition method based on deep learning of visible image is proposed.