Fractal analysis has been applied as a useful tool to quantitatively evaluate the streamer structure in insulating liquids. However, a fixed global greyscale threshold in the image binarization process can cause inevitable background noises and increase the uncertainty of the fractal analysis. This paper focuses on enhancing the fractal analysis by developing a dynamic greyscale threshold algorithm. The greyscale threshold for each pixel is dynamically estimated for a more accurate image binarization process. A unique background recognition method composed of two critical greyscale values is proposed to further reduce background noise in the binary streamer image. From the sensitivity study done in this paper, the square width in the algorithm was optimized at 80 pixels, while the difference between the two critical greyscale values for background recognition is set in the range of 25-40. The dynamic greyscale threshold algorithm is successfully applied to images of negative streamers obtained in five insulating liquids to enhance the fractal analyses.
|Title of host publication||IEEE Conference on Electrical Insulation and Dielectric Phenomena 2020|
|Publication status||Accepted/In press - 27 Aug 2020|