Abstract: Nowadays, image processing is an important area of application of modern computer technology. The following image processing tasks are known such as filtering and restoring images, image segmentation, as information compression tools. The problems of image recognition, in addition to the classical task of recognizing figures of a given shape in an image, pose new problems of recognizing lines and angles in an image, recognizing the edges of an image. In this research work, the authors solved the problem of localizing text in an image. To recognize text in the image, it is necessary at the initial stage to localize the image area containing text characters. This task is relevant in various fields: in the analysis of television broadcasts, the search for information in media outlets, administrative services in banks, insurance companies, etc. The article describes a technique optimized by the authors for detecting text areas and highlighting characters in images taken from metal surfaces on which some text is laser-engraved. A specialized algorithm for processing (cleaning) an image received from a USB camera has been developed. The algorithm is based on methods of mathematical morphology. The purpose of the work was to create (improve) the methodology and develop a program module that dynamically determines the text area of the inscription deposited on a metal surface by laser engraving. Because of the study, an optimized algorithm for the preliminary processing of text images taken from metal surfaces using a USB device was obtained. The practical significance of this algorithm is determined by the use in the software developed for the industrial enterprise Komsomolsk-on-Amur aircraft factory as a software module for image pre-processing before recognition.
Index terms: image processing, mathematical morphology, laser engraving