Application of Machine Vision Algorithms in Geolocation Determination

Currently, there are no mass solutions for navigation in case of weak communication signal, for example, in the subway. Therefore, the purpose of this study is to develop a software component that determines the location within the subway using computer-aided vision algorithms with a dictionary, the accuracy of which is up to one station. First, existing solutions for navigation in the conditions of weak signal were reviewed. Possible options for using computer vision for these purposes were considered. As a result, the option of determining the location within the subway by recognizing textual information from a photo was chosen. After that, the required theoretical basis was formed. The notion of an image, its basic properties and classes was introduced. The review of existing methods of image processing was carried out: conversion of color images into halftone ones, alignment of image lightness, binarization. Also, the notion of image noise, its classification, as well as methods against it were introduced. In conclusion of the formation of the theoretical base, the existing methods of automatic character recognition from an image were investigated. The practical part was devoted to the design and testing of a software component implementing the chosen approach. At the first stage, a top-level description of its working principle was formed. Then the available theoretical basis was added by the items concerning the processing of lines and fuzzy search. Using it and the upper-level description of the component working principle, variants of algorithms for its implementation were designed. After that, all of them were tested to determine the most successful of them. Thus, as a result of the research, a component that determines the location within the subway with an accuracy up to 47% using machine vision with a dictionary was developed.