Toward a Comparative Study of Lane Tracking using Omni-directional and Rectilinear Images for Driver Assistance Systems

With a panoramic view of the scene, a single omnidirectional camera can monitor the interior and exterior of the vehicle simultaneously. Omni-directional cameras have been shown to produce useful data for determining head gaze orientation from within a vehicle. We examine the issues involved in integrating lane-tracking functions using the same omni-directional camera. We present analysis on the impact of the omni-directional camera’s reduced image resolution on lane tracking accuracy, as a consequence of gaining the expansive view. This was done by first presenting Omni-VioLET, a modified implementation of the vision-based lane estimation and tracking system (VioLET), and evaluating the performance of both lane-trackers by comparing their performance with ground-truth generated with images synchronously captured with the other images along the same freeway course. The results are surprising: With 1/10th the number of pixels representing the same space and about 1/3rd the horizontal image resolution as a rectilinear image of the same road, the omni-directional camera implementation results in only twice the amount of mean absolute error in tracking the left-lane boundary position. With 2.3 times the resolution of the rectilinear camera, an omni-directional camera is shown to be just as accurate (in terms of average spatial error) as the rectilinear version.