Insulators Extraction In Aerial Image Of Inspecting Transmission Line

Exposuring in the wild, insulators requiring regular inspection. But transmission lines are going through complicated region, like mountains, plains elc. Using traditional manual inspection, not only the heavy workload but also difficult conditions. During the "Eleventh Five-Year" period, the helicopters are widely used in the transmission line patrol.Meanwhile how to use computer vision technology and inspection data to identify the defects in the transmission lines, become the focus of the study.This paper mainly studies the insulators segmentation and identification of aerial images, which provided the conditions for post-insulator fault identification.It has a great significance for the digital, intelligent of helicopters patrol transmission lines. First, the the insulators image database shoud be created. the outdoor image is susceptible to noise interference, so it should be pretreated. Preprocessing part includs denoising and gray linear transform. denoising part uses the mean filter, median filter and stationary wavelet transform threshold (SWT) combined with median filtering mixed denoising three ways, and through The peak signal (PSNR) to evaluate the result.Then, the maximum variance between clusters (OTSU) has a large amount of calculation. on the basis of the analysis of the particle swarm algorithm, APSO was introduced. it can greatly improve the OTSU method’s efficiency. OTSU method, iterative threshold method, genetic algorithm (GA) combined with OTSU and APSO algorithm combined with OTSU applied for insulators segmentation, and the running time and the segmentation results were compared, the results show that APSO combined with the OTSU improved segmentation speed within an acceptable result.Last, using the generalized Hough transform(GHT) and mathematical morphology methods to identify the position of insulator. using template match, Generalized Hough transform can roughly identify insulator on the original image. Mathematical morphology eliminating the interference point or region of image after segmentation, extract the skeleton of image after expansion. according to the skeleton to determine the rectangle position. So as to achieve the purpose of identification.