Classification of Markers in the ARTool Kit Library to Reduce Inter-marker Confusion

Abstract

Fiducial markers are widely used in many Augmented Reality (AR) applications such as to manipulate AR objects, robot navigation, education and AR based games. ARTool Kit is an open source, easy to configure and well-documented tracking system which is widely used for designing marker based AR applications. The patterns inside the markers often cause inter-marker confusion in the library. Currently there is no specialized algorithm to ensure the markers distinction in the library and to reduce inter-marker confusion. In this paper we present algorithms for classification of the markers in the ARTool Kit library in a given number of distinct classes. The degrees of similarities among the markers are computed and stored in an N-by-N matrix and then kmeans algorithm is used to classify the matrix into k classes. It is recommended to use markers from different classes to reduce inter-marker confusion. The marker from a class of smaller size will have minimal confusion with others.