A novel lane feature extraction algorithm based on digital interpolation

This paper presents a novel lane edge feature extraction algorithm based on digital interpolation. Zooming towards the vanishing point of the lanes creates a visual effect of driving. The positions of the lanes should not change significantly on the image plane while the vehicle is moving along the lanes. Considering the position information, more accurate lane features can be extracted. A gradient based vanishing point detection algorithm is incorporated to select the zooming area. The proposed algorithm shows outstanding performance on extracting features belonging solely to the lanes from severe noise environment. The algorithm is capable of removing non-relevant features produced by shadows, other vehicles, trees, buildings etc. The extracted feature map was tested with a classic lane detection algorithm, used in LOIS system. The detection results show that the improved feature map is an important factor to the performance of the whole system.