Lane estimation by particle-filtering combined with likelihood computation of line boundaries and motion compensation

Roads are exposed to various kinds of noise, lighting conditions and weather; thus, robust lane localization is difficult. Our paper presents an algorithm for a probabilistic estimation of lane information and solves the problem by the combining particle-filtering (PF) with likelihood computation of pixels on line boundaries using Gaussian-like functions. Additionally, because a pitch or an abrupt yaw motion of the camera makes lane estimation imprecise, motion compensation is added to the estimation. Thus, the algorithm provides precise lane information to a driver assistant or autonomous driving system. The results of experiments show that the algorithm is well adapted to various lane conditions.