A Sparse Gaussian Approach to Region-Based 6DoF Object Tracking

Abstract

We propose a novel, highly efficient sparse approach to region-based 6DoF object tracking that requires only a monocular RGB camera and the 3D object model. The key contribution of our work is a probabilistic model that considers image information sparsely along correspondence lines. For the implementation, we provide a highly efficient discrete scale-space formulation. In addition, we derive a novel mathematical proof that shows that our proposed likelihood function follows a Gaussian distribution. Based on this information, we develop robust approximations for the derivatives of the log-likelihood that are used in a regularized Newton optimization. In multiple experiments, we show that our approach outperforms state-of-the-art region-based methods in terms of tracking success while being about one order of magnitude faster. The source code of our tracker is publicly available.