Research On Fault Diagnosis Method Of Power Equipment Based On Infrared Image Feature Analysis

Infrared detection technology can effectively monitor the thermal state of power equipment,and has been widely used in power equipment detection and thermal fault diagnosis.With the continuous development of smart grids,infrared images have increased dramatically,while traditional infrared fault detection relies on manual troubleshooting or manual feature extraction.The detection efficiency is low and the dependence on personnel experience is high.There is a problem that the ability to analyze and judge defects is uneven.Therefore,the intelligent development based on infrared image feature analysis will improve the detection level of infrared diagnostic technology,realize efficient and intelligent detection of infrared images,and ensure the safe operation of the power grid.This paper analyzes the characteristics of infrared images,uses convolutional neural networks to implement fault diagnosis of power equipment,and proposes an R-FCN and Open CV-based power equipment fault detection algorithm model,and improves the detection and recognition of small target power equipment in infrared images.Ability to improve the accuracy of fault diagnosis and determine the fault level:First,analyze the target detection method and detection mechanism based on convolutional neural network.Through analysis and comparison,the R-FCN network is selected as the main network part of the fault diagnosis model in this paper,which provides the necessary theoretical basis for subsequent research.Secondly,in order to reduce the noise in the image,an image denoising method combining adaptive median filtering and wavelet transform is proposed.The evaluation criteria are based on the signal-to-noise ratio,structural similarity and image visual effect.Compared with the noise method,the infrared image filtering of power equipment can be better achieved.In addition,the method of histogram equalization is used to enhance the infrared image to achieve the purpose of enhancing image contrast and reducing image distortion.Infrared image denoising and enhanced preprocessing can improve the accuracy of fault diagnosis and recognition of infrared images of power equipment.Then,in view of the problems of low segmentation accuracy,over-segmentation and inability to ensure the closedness of the edges in the existing image segmentation methods,an improved K-means image segmentation method is proposed,which can automatically detect the target and background detected in the image Phase separation,so that the image can reflect the true state of the target,to achieve the purpose of improving the accuracy of subsequent analysis and target detection.Finally,an improved R-FCN network combined with Open CV secondary diagnosis fault diagnosis model is proposed.In order to achieve accurate identification of small targets of power equipment,the residual module is optimized,and each residual block combines low-level features and high-level features.Train the feature extraction network to automatically extract the feature map of the sample fault,and use R-FCN to detect the fault area and fault level on the feature map.The test results are sent to Open CV for secondary diagnosis of the defect classification to further reduce the false alarm rate.Through specific simulation tests,the average accuracy of the improved R-FCN network for fault diagnosis of infrared images of power equipment has reached 80.76%,which is8.43% higher than that of the original R-FCN network.Through the performance test,the average detection accuracy is 84.6 %.The end-to-end process of feature extraction and fault detection and recognition is realized,which avoids the problem of single features due to manual extraction of fault features and the failure to effectively detect and identify faults in specific situations and scenarios.