Gray intensity image feature extraction of partial discharge in high-voltage cross-linked polyethylene power cable joint

Abstract

SUMMARY A feature extraction method for gray intensity image of partial discharge (PD) is applied to recognize the insulating defects in high-voltage cross-linked polyethylene power cable joint. The method is based on a two-directional and two-dimensional (2-D) maximum margin criterion (MMC). A 2-D orthogonal projection of gray intensity image of PD was performed in horizontal and vertical directions. Projected image data were taken as discriminant vector of different gray intensity images to solve the high dimensional and small sample size of PD gray intensity image. The nearest neighbor classifier was used to classify the PD gray intensity image to recognize different insulating defects in the joint. The recongnition results of four typical insulating defects in the laboratory indicated that the extraction speed of the gray intensity image feature and recognition rate of insulating defects are superior compared with the common principal component analysis and Fisher discriminant analysis with MMC. Copyright © 2012 John Wiley & Sons, Ltd.