Edge extraction method study based on maximum entropy for linear lane identifying and tracking

Abstract

In order to better abstract lane mark edge and identify it, this paper proposes a new edge extraction method based on maximum entropy. This method combines both one-dimension and two-dimension entropy information. Meanwhile, image window variation technology is also applied for lane mark edge extraction and lane mark parameters can be acquired based on the bi-normalized adjustable template. Finally lane mark real-time tracking is realized by applying trapezia AOI method.