A Method for Automatic Pole Detection from Urban Video Scenes using Stereo Vision

Abstract

Pole-like structures such as the ones used for traffic lights, traffic signs, utility poles, lampposts or even trees are encountered everywhere in urban scenarios. Because they are robust landmarks, they can help solve problems from the autonomous driving domain, such as localization, mapping, and navigation. In this paper, we propose a method that extracts poles from stereo camera information. First, the intensity images are analyzed to find areas of interest that could contain the desired landmarks. Then, we build U- and V-disparity maps that are used to estimate the position of the poles on the road images. Finally, we cluster the candidate regions of interest, which are then further refined to eliminate outliers. We also use an algorithm for enhancing the illumination of nighttime images, so that we can detect the desired landmarks at different times of the day. Our system is able to extract poles from the same road, on different driving conditions, days, or lanes, it accounts for the possibility of occlusions, and we are able to obtain both a relative and an absolute localization.