Research on bead image recognition algorithm for determination of insulator levers

Hydrophobicity of insulator surface is an important index to evaluate electrical properties of insulation material, and insulator level is decided by hydrophobicity of insulator surface. In this paper, we give the whole process of bead image recognition algorithm for the assessment of insulator level. The key technique is the effective segmentation for original hydrophobicity image. Firstly, in order to eliminate horizontal stripe noise in original images, image preprocessing should be done, through analysis and comparison, directional filtering is used. Secondly, Hough transform is used for detection of half-torus region, which are shadows of beads because of illumination, and apparently belongs to image background, and thus the regions are marked as background for the following segmentation. Thirdly, algorithm for region segmentation based on gray similarity and gradient similarity is used, which fully use features of bead image, it start from gray similarity and gradient similarity of inner bead regions, and through region growing, segmentation of bead regions is effectively finished. Finally, shape factor method combined with the biggest bead area method is used to determine hydrophobicity level. The experiments show that satisfactory results have been achieved for determination of insulator level.