Vehicle recognition and tracking from road image sequences

In this paper, synergies between the recognition and tracking processes for autonomous vehicle driving are studied. Object recognition is periodically performed to focus attention on specific parts of the visual signal and to assign them symbolic meanings. Tracking is used to maintain correspondences between objects identified at successive recognition instants as well as to provide further features (e.g., spatio-temporal trajectories) on which to base object-pose estimation. Results obtained by using complex road scenes are reported, which demonstrate the validity of the approach in terms of robustness, accuracy, and time responses.