Optimal orientation estimation for mobile robot in urban area

A vanishing point is a point in a perspective drawing to which parallel lines appear to converge. In this paper, a novel optimal robot orientation estimation method in urban areas is proposed. This method is based on the vertical vanishing point in one view and the horizontal vanishing point correspondences across two views. To obtain an optimal estimation result, an error-aware vanishing point estimation method is developed, leading to a minimum variance solution. By analyzing the error propagation in computing vanishing point, we convert the minimum variance estimation problem to a convex optimization problem, which is well studied in operations research. An orthogonality check is proposed for each candidate horizontal vanishing point. Then, a vanishing point matching method is developed to find the correspondences. Finally, some physical experiments are carried out to validate the proposed method.