Fast lane tracking for autonomous urban driving using hidden Markov models and multiresolution Hough transform

Kaplan, Kemal; Kurtul, Caner; Akın, H. Levent

doi:10.1108/01439911011037677

Cited by 5 publications

(1 citation statement)

References 14 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…Another drawback of existing vehicle lane detection methods is that they require the prior knowledge such as camera's parameters [11], [24] and scene's layout [25], [26]. However, this information is often not available in assistive navigation systems for blind people.…”

Section: Related Workmentioning

confidence: 99%

Pedestrian Lane Detection for the Visually Impaired

Phung

Bouzerdoum

2012

2012 International Conference on Digital Image Computing Techniques and Applications (DICTA)

View full text Add to dashboard Cite

Traveling safely in outdoor environments is one of the most challenging activities for vision-disabled people. To improve the mobility of these people, an assistive navigation system is necessary. This paper is a step towards developing this system. We propose a robust method to detect the pedestrian marked lane at traffic junctions. The proposed method involves three stages: patches of interest (POI) extraction, lane marker detection and lane detection. The POI extraction is first performed to detect image patches located on the lane marker boundaries using template matching. Potential lane markers are then formed by using shape and intensity information. Finally, the walkable lane region is determined by verifying pairs of detected potential makers. A probabilistic framework using multiple geometric cues is proposed for the verification step. We create a new dataset collected from diverse environments and evaluate the proposed method on this dataset. Abstract-Traveling safely in outdoor environments is one of the most challenging activities for vision-disabled people. To improve the mobility of these people, an assistive navigation system is necessary. This paper is a step towards developing this system. We propose a robust method to detect the pedestrian marked lane at traffic junctions. The proposed method involves three stages: patches of interest (POI) extraction, lane marker detection and lane detection. The POI extraction is first performed to detect image patches located on the lane marker boundaries using template matching. Potential lane markers are then formed by using shape and intensity information. Finally, the walkable lane region is determined by verifying pairs of detected potential makers. A probabilistic framework using multiple geometric cues is proposed for the verification step. We create a new dataset collected from diverse environments and evaluate the proposed method on this dataset. The experimental results under challenging conditions have illustrated the efficiency and robustness of the proposed method compared to conventional techniques.

show abstract