Research On Infrared Image Processing In Insulator Contamination Detection

With the development of industry in our country, security and stability of power system is more and more important. Flashover of insulators is dangerous to the reliability unlimited. It seriously affects the security of the power system. In order to decrease the flashover accidents, it is especially important to detect the contamination level of insulators accurately. Infrared image detection method has been used in fault diagnosis and detection of power equipment widely. In the method of insulator contamination severity detection based on infrared image, image processing is very important. In this paper, it has done some research to contaminated insulators'infrared image processing methods.With low SNR (signal-to-noise rating) characteristics of insulator infrared images, it uses DT-CWT (dual-tree complex wavelet transform) and LAWML (Locally Adaptive Window-based de-noising using Maximum Likelihood) estimation to give an adaptive de-noising method to infrared images. It adopts LAWML to estimate standard variance of the real signal wavelet coefficient. It improves the accuracy and adaptability of the method. From the results of image processing, the method has good de-noising effect. It improves SNR of the image. In the same time of de-noising, it ensures the integrality of the image.In order to separate insulator and background, it brings forward a method based on window gray extension, which is easy to carry out. After image enhancement with the window gray extension, it adopts OTSU to confirm the threshold of binary image. With the segmentation problem of insulator and lead which was caused by few temperature difference between them in the infrared image, it uses the way based on gradient edge contour scanning to wipe off the lead, achieves accurately segmentation.In insulator string infrared images, there are barrier phenomena between adjacent insulators. In order to obtain comparable and real reflection of the contamination situation porcelain cup domain image, this paper adopts least square algorithm to fit eclipse with porcelain cup edge. Then it extracts half porcelain cup with the boundary of eclipse's major axis. Experiments results indicate that this method can extract object we interested: half porcelain cup successfully.With Algorithms mentioned above carried out successfully, it improves the accuracy of image analysis and settles a basement for contaminated level's correct estimation.