为准确评估输电线路覆冰程度以保证电力系统安全稳定运行,研究了运行绝缘子自然覆冰的图像分割方法,提出了覆冰程度评估参数—覆冰绝缘子面积变化百分比。对比研究了阈值法、Sobel、Laplacian和Canny算子法、种子区域增长法和Grab Cut图论法等6种图像分割方法对田地、树木和丘陵等3种典型背景和未覆冰、轻微覆冰、覆冰和严重覆冰等4种覆冰程度的覆冰绝缘子分割效果,结果表明与其他方法相比,Grab Cut图论法可以更精确地分割出覆冰绝缘子。利用该方法分割南方电网输电线路灾害(覆冰)预警系统2014–12—2015–02的10组典型运行绝缘子覆冰图像,计算得到覆冰绝缘子面积变化百分比,并与人工观冰标准厚度和系统称重法监测的等效覆冰厚度进行了对比研究,结果表明,覆冰绝缘子面积变化百分比与人工观冰标准厚度的关联性达0.79。
In order to guarantee the safety and stability of power system, we investigated image segmentation methods, and proposed a characteristic parameter, namely, icing-area change percent, for evaluation of natural icing condition on in-service insulators. Six segmentation methods, including threshold, Sobel, Laplacian, Canny, seed region growth, and graph-theory-based GrabCut, were investigated for natural icing on in-service insulators. The cases cover three typical backgrounds of fields, trees, and hills, and four icing conditions of free of icing, light icing, icing ,and serious icing. The results show that the GrabCut method performs the best accuracy for all the experiment cases. Based on this method, ten groups of typical images from China Southern Power Grid transmission line disaster (icing) warning system from De- cember 2014 to February 2015 were segmented and calculated for icing-area change percent. Then, the icing-area change percent was compared with the standard icing thickness by manual observation and the equivalent icing thickness by sys- tematic weighing method. The results show that icing-area change percent has a similar trend to the standard icing thickness by manual observation, with a correlation coefficient of 0.79.