Model-based vision techniques originally developed for the recognition and pose recovery of a vehicle in a single image, are used here to track a vehicle through a sequence of images. The knowledge of the position of the camera with respect to the ground plane is used to reduce the search space of possible vehicle positions from six dimensions to three.

Model-based vision allows the recovery and tracking of the 3D position and orientation of a known object from a sequence of images. A Kalman filter can be used to improve the tracking stability with three main benefits. Firstly it is an optimal filter in the least squares sense, with the added advantage that the physical dynamics and constraints of the tracking problem can easily be built into the system model. Secondly the measurement model allows for uncertainty in the measurement of the recovered model position. Thirdly, data on the empirical error are generated which can, for example, be used to control the model matching process used in the tracking.