Serhiienko A. Models and method of information technology for identification of hot-metal transfer ladle number by its noisy photo image

The thesis is devoted to the solution of an urgent scientific and practical problem of developing new information technology to identify the number of a hot-metal transfer ladle by its noisy photo image. It is proposed and substantiated to use the median filtration method, which is optimized to speed, at the first stage of the preliminary processing for the first phase (filtration). The method of local threshold global binarization of the image by automatic calculation of the threshold value based on its average brightness, which is justified to use for the second phase of the first stage of the preliminary processing (binarization), has been improved. For the localization of the information area (the second stage of processing), it is proposed to perform the segmentation based on image clustering by searching for the nearest neighbor with a modification of the search using a sliding window. To carry out the third stage of processing, identification, the model and method of comparison of the localized image with the reference one based on the image indistinct characteristic have been developed. It considers the different size of numbers put on ladles, different illumination intensity of input photo images, numbers put manually without a stencil, numbers with noise in the form of pieces of cast iron, stains of lime, a large amount of dust, etc. An example of using the proposed information technology in production conditions is given. It is researched that the use of the method of comparison of the localized image with the reference one based on the fuzzy image characteristic allows to perform the identification with high-level reliability. The proposed information technology is implemented in the system of monitoring the movement of hot-metal transfer ladles and the condition of cast iron.