Advanced Road Vanishing Point Detection by Using Weber Adaptive Local Filter

Abstract

Variations in road types and its ambient environment make the single image based vanishing point detection a challenging task. Since only road trails (e.g. road edges, ruts, and tire tracks) would contribute informative votes to vanishing point detection, the Weber adaptive local filter is proposed to distinguish the road trails from background noise, which is envisioned to reduce the workload and to eliminate uninformative votes introduced by the background noise. This is possible by controlling the number of neighbors and by increasing the sensitivity for small values of the local excitation response. After road trail extraction, the generalized Laplacian of Gaussian (gLoG) filters are applied to estimate the texture orientation of those road trail pixels. Then, the vanishing point is detected based on the adaptive soft voting scheme. The experimental results on the benchmark dataset demonstrate that the proposed method is about 2 times faster in detection speed and outperforms by 1.3% in detection accuracy, when compared to the complete texture based gLoG method, which is a well-known state-of-the-art approach.