Insulator Image Segmentation Based On Non Subsampled Contourlet Transform

Insulator is an indispensable element on transmission lines with the dualfunction of wire support and electrical insulation. And frequent faults of insulator,such as filth, crack, and breakage affect the normal operation of transmission linesand shorten the operating life of lines. Even worse, some faults may lead tolarge-scale blackout and huge losses. Therefore, it is necessary to real-time monitorrunning status of insulator. Helicopter inspection is an important patrol way ontransmission lines with the advantage of efficiency, reliableness and saving resource,and its rapid development adapts to the development of smart grid in our country.Automatic segmentation and localization of insulator in aerial image is theprerequisite of condition monitoring and fault diagnosis of insulator.Aiming at aerial image with the characteristics of complex background and lowresolution, this paper proposes a threshold segmentation method based on NonSubsampled Contourlet Transform (NSCT). First, construct gray entropy model basedon NSCT decompose. Secondly, calculate threshold by Bacterial Foraging-ParticleSwarm Optimization (BF-PSO) algorithm. Finally, segment the insulator image bythreshold and obtain a binary image with the separation of foreground andbackground.On the basis of binary image including insulator, towers and power lines, thispaper proposes a localization method base on binary shape features of insulator string.And this method digitally describes and makes full use of the shape features in binaryimage, removes non-insulator-targets and keeps information of insulator by thedescriptions, and automatically localizes insulator string in aerial image.To verify its performance, compare the proposed method with classical methodsusing real aerial images. Experimental results show that the proposed segmentationmethod has a better anti-noise capability, and can separate target with complete edgesand little internal holes from background; the location method can also realize thelocalization of insulator in complex background, with accurate positioning results,fast running speed and no artificial participation that can lay the foundation for statemonitoring and fault diagnosis of insulator.