Navigating safely in outdoor environments is a challenging activity for vision-impaired people. This paper is a step towards developing an assistive navigation system for the blind. We propose a robust method for detecting the pedestrian marked lanes at traffic junctions. The proposed method includes two stages: regions of interest (ROI) extraction and lane marker verification. The ROI extraction is performed by using colour and intensity information. A probabilistic framework employing multiple geometric cues is then used to verify the extracted ROI. The experimental results have shown that the proposed method is robust under challenging illumination conditions and obtains superior performance compared to the existing methods. © 2012 ICPR Org Committee.