期刊名称：IAENG International Journal of Computer Science
印刷版ISSN：1819-656X
电子版ISSN：1819-9224
出版年度：2019
卷号：46
期号：2
页码：300-310
出版社：IAENG - International Association of Engineers
摘要：Damaged insulators may affect the mechanical andelectrical performance of an electric power grid, which canlead to the flow of leakage currents through the line supports.This increases electrical losses and voltage drop in the powergrid. It also poses a risk to human safety. Thus, it is crucialto monitor and inspect the condition of insulators to detectdegradation or damage. However, the traditional method ofinspection is inadequate in meeting the growth and developmentof the present electric power grid. Hence an automatedsystem such as the computer vision method is presently beingexplored as a means to resolve this crisis safely, speedily andaccurately. This paper presents a method that distinguishesbetween defectuous and non-defectuous power-line insulators.Active Contour model is applied for insulator segmentation inorder to determine insulator region of interest. Local binarypattern is used for feature extraction from the insulator regionof interest which is then fed to the support vector machineclassifier for classification. An accuracy of 94.1% was achievedwhen morphological operation is used in combination withactive contour model for segmentation based on the groundtruths.In addition, local binary patterns feature extractionmethod outperformed gray level co-occurrence matrix whenused with support vector machines.
关键词：Active Contour Model; Local Binary Patterns;Power-line Insulator; Support Vector Machines
