Damper Recognition And Location For Transmission Line Aerial Image

Nowadays, the new energy reform with characteristics including clean energy and smart grid are worldwide boosted, intelligent inspection and defect detection of transmission lines are the important parts of the smart grid. During the 11th five-year plan, the inspection technology of transmission line by helicopter has been widely applied. But the target recognition and the fault detection for the transmission lines of aerial images are still not completely out of the artificially diagnostic pattern. How to defect fault for transmission lines components accurately and efficiently is still a hotspot and difficult issue.This paper launches the research on damper identification of inspection transmission lines images by helicopter, which prepare for the defects diagnosis of dampers. Firstly, according to the relationship that dampers are the components of power lines, damper image database is constructed. Secondly, image processing including de-noising, image enhancement and image segmentation for database images are carried out. Mean filter,median filter and the combination of stationary wavelet transform and median filter are used to de-nosing images respectively:The peak value signal-to-noise ratio and the mean square error are used to evaluate the de-nosing effect; the gray contrast-enhanced transfom experiments is carried by analyzing of damper the gray characteristics of damper images; Iteration,OTSU,maximum entropy are used to segment images.Finally, the general Hough transform (GHT) method that is improved and special Hough transform method (SGHT) are used to identifying the images processed respectively. The two improved methods that are in the basis of classic GHT:One is Harris corner based on GHT. In this algorithm, instead of the contour edge character, Harris corner features is introduced to reduce the space complexity and the time complexity; the other is modified GHT. In this algorithm, the space domain in GHT is replaced by the frequency domain, according to the correlation of the fast Fourier transform (FFT), rotated and scaled template image is matched with the preprocessing image to be detected, in view of the actual voting peak gathering but not converged, merger peaks is proposed and used to extract peaks by the means of peak limited by threshold and designating the peak number. Compared with Harris corner based on GHT,SGHT and modified GHT, the modified GHT is the best algorithm for recognizing and locating damper. What's more, merge peaks by designating the peak number has a higher correct rate and a lower error rate of detection.