Failure Diagnosis Of Infrared Images Of Power Equipment Based On Improved YOLOv4

With the increasingly maturity of infrared technology and the rapid development of target detection algorithms,these two technologies are deeply integrated,which will be able to solve many problems in real life.In order to realize the real-time online monitoring of power equipment failure,solve the low efficiency,improve the fault diagnosis level of power equipment,study the applicability of different target detection methods to real-time detection,studied how to expand the data structure more perfect,studied how to judge the thermal failure level of power equipment according to the temperature of infrared image,and studied how to make GUI interface to show the research content.First,for the problem of scarce directly available datasets,the web search,book books,and the collected data set uses image processing related technology to expand the data set through rotation,stretching and adjusting image parameters.Only power equipment insulators,switch and transformer are included in the dataset.Some images also need to be pre-processed before labeling data sets,including gray scale histogram equalization and simple filtering processing,which can improve the quality of images and thus provide a good data basis for real-time detection of power equipment.Secondly,in order to solve the balance between real-time and detection accuracy,some existing target detection algorithms are needed and improved.By improving the YOLOv3 algorithm,the average accuracy was improved,but the detection speed remained in the original place.Then,by improving the YOLOv4 algorithm,the average accuracy of the improved YOLOv4 algorithm reached 91.23%,and the detection speed reached 69.1fps,is greatly improved compared to other algorithms.Finally,for the improved power equipment target detection algorithm YOLOv4 combined with the existing power equipment infrared image heat fault judgment method,the identified power equipment uses the surface temperature difference judgment method to mainly judge the fault level for the current heat generating device.The GUI interface made by MATLAB software,through the layout of the control and callback function,the identification and labeling of selected picture power equipment can be realized,while using the fault level and surface temperature difference judgment to display the maintenance suggestions.The experiment show that the fault diagnosis system designed in this paper can determine the real-time detection and fault level of power equipment.Figure 42 Table 8 Reference 66...