Markerless Tracking for Augmented Reality Using Different Classifiers

Autor:	Mohiy M. Hadhoud, Faten A. Khalifa, Noura A. Semary, Hatem M. El-Sayed
Rok vydání:	2015
Předmět:	Support vector machine ComputingMethodologies_PATTERNRECOGNITION business.industry Computer science FREAK Augmented reality Computer vision Artificial intelligence Bayes classifier Feature description business Classifier (UML) Random forest
Zdroj:	Advances in Intelligent Systems and Computing ISBN: 9783319266886 AISI
DOI:	10.1007/978-3-319-26690-9_3
Popis:	Augmented reality (AR) is the combination of a real scene viewed by the user and a virtual scene generated by the computer that augments the scene with additional information. The user of an AR application should feel that the augmented object is a part of the real world. One of the factors that greatly affect this condition is the tracking technique used. In this paper, an augmented reality application is adopted with markerless tracking as a classification task. ORB algorithm is used for feature detection and the FREAK algorithm is used for feature description. The classifiers used for the tracking task are KNN, Random Forest, Extremely Randomized Trees, SVM and Bayes classifier. The performance of each classifier used is evaluated in terms of speed and efficiency. It has been observed that KNN outperforms other classifiers including Random Forest with different number of trees.
Databáze:	OpenAIRE
Externí odkaz:	https://explore.openaire.eu/search/publication?articleId=doi_________::d2591439bf7493422d8215e8feff424c https://doi.org/10.1007/978-3-319-26690-9_3 Zobrazit plný text záznamu