Portal de Periódicos da CAPES

Artigo Produção Nacional

BIBTEX RIS

RECOGNIZING AND EXPLORING AZULEJOS ON HISTORIC BUILDINGS’ FACADES BY COMBINING COMPUTER VISION AND GEOLOCATION IN MOBILE AUGMENTED REALITY APPLICATIONS

2017; Academy Publisher; Volume: 13; Linguagem: Inglês

10.5555/3177197.3177201

1796-2048

G R SantosCarlos, D O AraújoTiago, R ChagasPaulo, Nelson Neto, Bianchi Serique Meiguins,

Spatial Cognition and Navigation

Mobile augmented reality (MAR) applications assist users in navigating and exploring their actual surroundings, displaying virtual contents that correspond to objects and scenes in the real world. However, despite the growing popularity of these applications, some experiences can be frustrating when users are unable to correctly recognize Points of Interest (POI), objects, or places they want to visit or obtain more information. The misleading recognition can occur due to imprecise Global Positioning System (GPS) data or a lack of QR codes for interaction. Hence, this article presents a proposal that combines pattern recognition in images with geolocation information to improve the accuracy of the identification of POIs. The usage scenario is the identification of azulejos (tiles) on the facades of historic buildings in the city of Belem of Para, Brazil. This issue is relevant based on similarities between azulejos and its huge amount of different types, whose variety of designs and colors of geometric forms can make the identification a hard task. The used methods to extract the azulejos’ features were the co-occurrence matrix combined with color percentage, and the global positioning data to increase the accuracy of classification because similar azulejos can be geographically far apart. Tests were conducted using six machine learning algorithms (neural network, decision tree, k-nearest neighbors, naive Bayes, random forest, and support vector machine) of different paradigms. The first results show that the pattern recognition in images combined with geolocation information is a promising approach for better identification of the POIs in MAR applications.