Geometrically constrained object tracking in non-overlapping calibrated cameras within a bayesian framework

Abstract

When establishing correspondence between objects across non-overlapping cameras, the existing methods combine separate likelihoods of appearance and kinematic features in a Bayesian framework, constructing a joint likelihood to compute the probability of re-detection. So far, no method has assumed dependence between appearance and kinematic features. In this work we introduce a novel methodology to condition the location of an object on its appearance and time, without assuming independence between appearance and kinematic features, in contrast to existing work. We characterize the linear movement of objects in the unobserved region with an additive Gaussian noise model. Assuming that the cameras are affine, we transform the noise model onto the image plane of subsequent cameras. This noise model acts as a prior to improving re-detection. We have tested our hypothesis with toy car experiments and real-world camera setups. The prior constrains the search space in a subsequent camera, greatly improving the computational efficiency. Our method also has the potential to distinguish between similar-type objects, and recover correct labels when they move across cameras.