Search

The article describes the information technology for recognition of the number on the hot metal transfer ladle. Processing stages have been identified and described: preliminary processing, localization and number identification. The filtration and incomplete thresholding by means of global binarization with an automatic choice of a threshold have been analyzed. Localization is executed by a single linkage (the nearest neighbour) method with the modified windowing search algorithm. After localization area has been identified, a segment is generated. Identification is accomplished by means of a unique method, which is based on an indistinct representation of images and their comparison with indistinct standards. Combining in a characteristic point the segment which is being analyzed with a prototype segment kernels are formed. Those are then fractured into blocks. Based on these blocks two fuzzy sets are obtained -- one for the segment that is being analyzed and one for the prototype. For these sets the relative Hamming distance is then calculated. The calculated value shows the degree of similarity of the segments. During the tests it was proved that the Hamming distance for the segments with identical numbers is 0,01÷0,16, whereas for the segments with numbers of different ladles it is 0,26÷0,49, which is enough for a successful identification of a ladle by the image of its number. [ABSTRACT FROM AUTHOR]