Robust multi-hypothesis 3D object pose tracking

Abstract

This paper tackles the problem of 3D object pose tracking from monocular cameras. Data association is performed via a variant of the Iterative Closest Point algorithm, thus making it robust to noise and other artifacts. We re-initialise the hypothesis space based on the resulting re-projection error between hypothesised models and observed image objects. This is performed through a non-linear minimisation step after correspondences are found. The use of multi-hypotheses and correspondences refinement, lead to a robust framework. Experimental results with benchmark image sequences indicate the effectiveness of our framework. © 2013 Springer-Verlag.