Recognition and Application of Infrared Thermal Image Among Power Facilities Based on YOLO

This paper researches the infrared thermal image recognition and appliacation of power facilities based on YOLO neural network. The preliminary work includes the construction of the power facilities data set, and the preprocessing of the photo noise reduction. After sending the infrared thermal image to the YOLO neural network, the system uses the Bounding Box to crop out all possible electrical equipment and names the device. Then a nonlinear least squares curve is used to measure the highest temperature of the device. Testing on different devices such as Combine Filter, Porcelain Sleeve and Isolation Switch shows that the system can accurately and stably identify power facilities. The least squares curve can accurately locate the highest temperature of the device. The system can effectively reduce labor costs and achieve high recognition accuracy.