An extended hyperbola model for road tracking for video-based personal navigation

We present a robust road detection and tracking method using multiple vanishing points and the condensation filter. We represent the road using an extended hyperbola model with an added non-linear term to handle transitions between straight and curved road segments. The parameters of the road model are estimated using multiple vanishing points located in road segments. A vanishing line is then determined using a robust iterative curve fitting technique to recover parameters of the road model. These are then fed into a robust condensation tracker [M. Isard, A. Blake, Condensation: conditional density propagation for visual tracking, Int. J. Comput. Vision, 1998] to track the road. The tracker is able to deal with difficult road conditions. Experiments using real road videos demonstrate the suitability of our approach for real-time applications. A comparison with the Kalman filtering technique demonstrates the robustness of our approach.