Accuracy Analysis of Augmented Reality Markers for Visual Mapping and Localization

Abstract

An important problem that can be solved by Computer Vision algorithms is computing a 3D reconstruction of a scene captured by an ordinary camera. While some approaches seek solutions based on natural features (projections of the captured scene in the image), artificial markers can be used as an attractive alternative because they are easier to detect and less susceptible to error. Several artificial markers systems are available as computer libraries with different approaches of camera pose estimation, although their capabilities are still unexplored by Simultaneous Localization and Mapping (SLAM) solutions. This work presents a solution to the problem of SLAM using artificial markers. The automatically computed camera poses by the ARUCO library are assessed on data depicting commonly encountered scenarios in Robotics and Augmented/Virtual Reality applications, such as small to medium sized rooms and a research laboratory. The evaluation considers an established criteria that should clarify the accuracy of the use of artificial markers in visual mapping and localization.