Research On Image Segmentation Of Insulators In Transmission Line By Patrol Inspection With Helicopter

The method of patrol inspection with helicopter has considerable advantages of efficiency, reliability, rapidness and disregard of distance. Images acquired in patrol inspection provide basis data for intelligent fault diagnosis. Nevertheless damaged insulators, as for the common occurrences, profound impact the safety operation of transmission line. Nowadays, image segmentation of insulators under the complicated background haven't been settled. This paper do the research on the project of image segmentation of insulators, furthermore, the result could provide basis for the determination of follow-up parts.This paper realistically explored every possible avenues and tactics on the merit of the characteristic of the insulator Image to propose the following approaches.As the beginning, we subdivided the insulator images into two types: the blue insulator image and the black insulator image.With the regard of the blue insulator image, we located the blue insulators by the means of tilt correct and projection, so that we could get the image included insulators only and excluded complicated background effectively. Therefore, we segmented the images of position-finding by the way of the saturation component in the HSI color space and obtained accurate outcome.This paper advanced the segmentation of the blue insulators with a new method: Firstly, we folded the hue photograph in order to release the false segment of the background in the selected HSI color space. Secondly, we segmented the folding hue photograph with the method of maximum variance between clusters. In accordance with the shortages of the segmentation image of the single component of H or S, we mixed the two segmentation outcomes, and got the final result by the method of morphology.We found that the saturation of the black insulators is low, so that we acquired the image of the complete black insulators with the method mixed with the saturation component in the HSV color space and maximum variance between clusters. With regard to other low saturation resemblance, we came up a solution with the principal component analysis, by which way we got the accurate outcome.All in all, this paper interpreted the methods we mentioned could provide ideal segmentation of insulators.