| abstract |
The invention discloses a method and a system for detecting temperature difference of metal surface of electrical equipment. Under the condition of reference temperature and the temperature rising condition of preset temperature rise step, a visible light image data set of metal surface of electrical equipment of metal material commonly used in electrical equipment is established, and the image is extracted. The RGB pixel features of the corresponding pixels in different temperature regions in the middle are combined into a temperature difference feature data set, and the temperature difference feature data set is compressed and reduced in dimension to obtain a low-dimensional temperature difference feature data set, which is input to the temperature difference recognition model based on machine learning algorithm for training, and the obtained The trained temperature difference identification model is used to detect the temperature difference in different areas of the same electrical equipment, which solves the high cost of the existing method of temperature rise fault detection for electrical equipment, and the measurement results are vulnerable to other infrared radiation. The interference of the source and the need to manually correct the measurement results lead to technical problems that limit the application. |