Static Obstacle Detection along the Road with a Combined Method

Abstract

A smart phone application capable of recognizing numbers of oncoming buses would be of great assistance to blind individuals. To facilitate bus route number reading, obstacles along the road should first be identified. This paper is concerned with identification of static obstacles comprising two processes: the first process involves road area detection and is addressed by applying a rotational invariant of the uniform local binary pattern via k-means clustering. Furthermore, an artificial neural network is employed to select a group of k-means that contains the road area. Next, the straight lines on the road are detected via Hough line transformation. Finally, the line selection step is used to define the road area boundary. The second process involves static obstacle detection and is addressed through segmentation, vertical projection of the road area boundary, and consideration of the vertically projected signal. The experimental results demonstrate a high performance of the proposed method with an F-measure of 0.912.