Lane Detection and Tracking Using a Linear Parabolic Model

All over the world, safety is an important issue for humans. Therefore, reducing road accidents and hence, saving lives of individuals has become a great research interest in the context of advanced driver assistance systems. Among the whole complex and challenging tasks of future road vehicles, road lane detection or road boundaries detection will hold a key role. Lane detection and tracking is an important topic in autonomous navigation, since the navigable region usually stands between the lanes, especially in urban environments. Several approaches have been proposed for lane detection, but Hough transform seems to be the dominant among all and in the current thesis it is employed for lane detection aims. A robust lane tracking method is also required for reducing the effects of the noise and achieving the demanded processing time. Herein, a lane tracking method using linear parabolic model as a practical method has been applied for following the lanes. Next to the detection process, detected boundaries in consecutive video frames are tracked through a linear- parabolic model. The linear part of the model is used to fit the near vision field, while the parabolic model fits the far field. Connected component method was utilized to improve the lane detection process which was of great importance for the curved part of the road. The data from Carnegie Mellon University directory were used for applying the hybrid method and then three of them namely run2a, run2b and May30-90 were simulated. Thesis also provides simulation results for two custom videos from FAMAGUSTA-NICOSIA road taken under different lighting conditions. These sequences are named ROAD-22 and ROAD-23in this study. The achieved results demonstrated the outstanding performance of the proposed hybrid method. For all the available samples, the average percentage values for correctly detecting and marking the right and left lines were 96.3% and 97.12% respectively. Keywords: lane detection; lane tracking; Hough transform; linear-parabolic model; connected component analysis.