基于投影统计与双曲线拟合的车道识别算法

提出了一种运用投影统计与双曲线拟合的车道线识别方法。首先对采集的图像进行预处理，利用车道线宽度等信息提取车道线的特征点;然后根据特征点的位置及方向角分布对消失点进行定位，并基于消失点位置提取近视场直道模型，再结合直道模型对当前车道进行补充;最后利用近视野车道直线信息和双曲线拟合，对存在弯曲车道的车道线进行车道线寻优。实验结果表明，提出的算法能有效检测不同曲率的车道线，所提出的车道检测方法准确率高，在复杂行驶环境下表现出较高的鲁棒性。

Lane Detection Algorithm Based on Projection Statistics and Hyperbolic Fitting

A lane detection algorithm was proposed based on projection and hyperbolic fitting. Firstly, the collected images were preprocessed, and the feature points of lane marks were extracted by the informations such as the widths of lane lines. Secondly, the vanishing points were located according to the positions and the direction angles of feature points, lanes of near field were extracted based on the vanishing points, and then supplemented the lane. Finally, lanes of near field information and hyperbolic fitting were used to find the best lanes where there were a crooked road. The experimental results show that the algorithm proposed herein may effectively detect the lanes with different curvatures, and the proposed lane detection method has high accuracy rate. The algorithm exhibits high robustness under various environments.