Automatic Calculation of Icicle Morphological Parameters for Multi-Perspective Suspension Insulators Using Gradient-Contour Analysis

Abstract

Glaze icing significantly increases the flashover probability of in-service insulators, making it the most hazardous icing type for overhead power lines. To automatically and accurately characterize the glaze icing state of suspension insulators susceptible to icicle bridging, this paper proposes a novel method for calculating icicle morphological parameters, including icicle distribution and bridging degree, using images from multi-perspective monitoring scenarios. The method first recognizes and segments images to focus on insulators, then uses gradient-contour analysis to correct axial tilt, locate shed units of different materials, and detect icicles between sheds. It utilizes the relative positions of shed units and icicles to quantify uneven and segmented icing phenomena and calculate the icicle bridging degree of the insulator string and shed units. Image analysis of 15 multi-view, multi-material, and varying-length suspension insulators demonstrates that this method achieves 98.8% precision in localizing 91 shed units and 94.6% precision in detecting 490 icicles. The accuracy in recognizing axial and circumferential icicle distributions is 91.2% and 98.9%, respectively, with precision surpassing 90% in calculating the icicle bridging degree. The method can finely parameterize icicles’ position and length, providing a solution for image-based diagnosis of distribution uniformity and bridging severity of glaze ice on insulators.