K-means indiscernibility relation over pixels
Peters, J.F., Borkowski, M. K-means indiscernibility relation over pixels. In: S. Tsumoto, R. slowinski, J. Komorowski, J.W. Grzymala-Busse, Eds. Rough Sets and Current Trends in Computing, LNAI 3066. Berlin, Heidelberg, Springer, 2004, 580-585
This article presents a new form of indiscernibility relation based on K-means clustering of pixel values. The end result is a partitioning set of pixel values into bins that represent equivalence classes. The proposed approach makes it possible to introduce a form of upper and lower approximation specialized relative to sets of pixel values. This approach is particularly relevant to a special class of digital images for power line ceramic insulators. Until now the problem of determining when a ceramic insulator needs to be replaced has relied on visual inspection. With the K-means indiscernibility relation, it is now possible to automate the detection of faulty cermic insulators. The contribution of this article is the introduction of an approach to classifying power line insulators based on rough set methods and K-means clustering in analyzing digital images.