Research On Segmentation And Recognition Method For Insulator Image Based On Complex Network Community

In order to ensure the safe and reliable operation of the w hole power system,the insulator must be repaired and maintained in time.Before this,it is important to recognize and extract the insulator from images with complex background.In this paper,the complex network community is introduced into the image segmentation and recognition,and proposed insulator segmentation and recognition methods based on complex network community.Designed an insulator image with simple background segmentation method based on super pixel and complex network community.The method can reduces the computation cost by using TP(Turbo pixel).According to the complex network community model and the selected feature,the insulator will be expressed by a separate community.As the experimental results shown,the propose method has obvious advantages in comprehensive quality Y value and intersection-over-union(IOU)value.The method can segment insulator target accurately.In order to segment aerial insulator images with complex background,especially images with pseudo target(such as transmission lines and towers)and insulator adhesion.This paper proposes an aerial insulator image with complex background segmentation method based on hybrid feature and complex network communit y.Firstly,the method removes the text,then uses TP to segment the image and calculates hybrid feature with features of each super pixel in a certain weight.Finally,use the adjacency that consisting of hybrid feature to segment the insulator segmentation into communities.The accuracy of this segmentation method is higher.It can remove most pseudo targets and retain the whole information of the insulator.In order to solve the insulator recognition problem,this paper use generalized Hough transform method to recognize the insulator community from the communities segmented by the complex network community method.Firstly,we process the segmentation results into binary image and then extract their edge information by canny edge detection algorithm.Then match the template information and the segmentation communities.Finally,we recognize the insulator community from the segmentation communities according to the shape feature.Our method see the internal information of image as nodes to construct image network.It can segment insulators from the simple background images or the complex background images or noisy images.Then match them with the template to recognize them.The proposed method is efficient,robust and accurate.