A lane detection algorithm based on reliable lane markings

This paper proposes a robust and effective vision-based lane detection approach. First, two binary images are obtained from the region of interest of gray-scale images. The obtained binary images are merged by a novel neighborhood AND operator and then transformed to a bird's eye view (BEV) via inverse perspective mapping. Then, gaussian probability density functions are fit to the left and right regions of a histogram image acquired from the BEV. Finally, a polynomial lane model is estimated from the identified regions. Experimental results show that the proposed method accurately detects lanes in complex situations including worn-out and curved lanes