Robust 3D marker localization using multi-spectrum sequences

Abstract

Robust 3D marker localization in different conditions is an open, challenging problem to stereovision systems. For years, many algorithms - using monocular or multiple views; based on visible or infrared sequences - have been proposed to solve this problem. But they all have limitations. In this paper, we propose a novel algorithm for robust 3D marker localization in different conditions, using synchronous visible and infrared (IR) spectrum sequences captured by binocular camera. The main difficulty of the proposed algorithm is how to accurately match the corresponding marked objects in multi-spectrum views. We propose to solve the matching problem by considering geometry constraints, context based features of special designed markers, 3D physical spacial constraints, and etc. Experimental results demonstrated the feasibility of the proposed algorithm. © 2009 Springer-Verlag.