Monocular detection of order, collision, and other kinematic events

For arbitrary motions with polynomial regimes in their translational component, the authors describe a framework for representing monocular visual information as a temporal map, organized according to temporal parameters. These parameters include the actual time-to-collision and time-to-synchronization calculated according to the detected model order. The authors discuss the recovery and estimation of these parameters and the model order and collision detection schemes. Experiments on real and simulated dynamic sequences are reported.