РАЗРАБОТКА АЛГОРИТМА РАСПОЗНАВАНИЯ ИЗОБРАЖЕНИЯ ПОСРЕДСТВОМ КОМБИНАЦИИ АЛГОРИТМОВ ПОИСКА ПО ФОРМЕ И ТЕКСТУРЕ

Abstract

The article discusses image’s recognizing methods by a contour, for an object’s color characteristics highlighting, and object’s recognizing by an various angles. Each of these methods has its own implementation features, advantages and disadvantages. The paper considers a search method for shape and texture characteristics through Tamura analysis, such as histograms of individual colors and texture depth. An algorithm is also considered for obtaining Fourier descriptors, which allows you to match the shapes of areas using graph matching methods. To describe the image’s contour when selecting an object from the background, the Grid method is used. This method allows you determination the image’s contour, and then search for this contour, comparing it with the standard. The Grid method is used to describe the image’s outline with it’s selection from the background. The work implements an image search for all three parameters simultaneously during to a methods combination that described above for errors reducing in the algorithm. This method will not depend on the original and recognized images angle, on the shooting quality difference between the original and recognized images. The image searching algorithm involves comparing the image with the samples, that stored in the database. Comparison with the sample can be made with determining the characteristics inherent in the product or products’ group. The paper presents of the resistance to errors study to the proposed method, the correctness of image’s recognition by it and also considers its speed. The article considers image’s recognizing methods by a contour, by an object’s color characteristics highlighting, and object’s recognizing from various angles. Each of these methods has its own implementation features, advantages and disadvantages. The paper considers a search method for shape and texture characteristics through Tamura analysis, such as histograms of individual colors and texture depth. An algorithm is also considered for obtaining Fourier descriptors, which allows matching the shapes of areas by graph matching methods. To describe the image’s contour when selecting an object from the background, the Grid method is used allowing to determine the image’s contour and then search for this contour, comparing it with the standard. In the paper, an image search using all the three parameters simultaneously is realized by combining the methods described above while reducing the errors in the algorithm. Such a method will not depend on the original and recognized images angle, on the shooting quality difference between the original and recognized images. The image searching algorithm involves comparing the image with the reference standards stored in the database. Benchmarking is done by identifying the characteristics inherent in the product or products’ group The paper presents a study in the resistance to errors of the proposed method, in the correctness of image’s recognition by it and also considers its processing speed.