Advanced lane line detection using OpenCV

After applying a perspective transform of the current viewpoint to a top-down view, channel combination in L*a*b* color space is applied to obtain a grayscale image, which is then fed to a template matching system which locates lane pixels and suppresses non-lane ones. A windowed search is instantiated given a histogram peak for each lane line (left and right), followed by a windowed search over along the candidate pixels. A second order polynomial is fit to the detections and backprojected to view space.