基于特征点提取的弯道识别算法研究

针对高速公路弯道处事故率高的特点,研究了一种基于特征点提取的高速公路弯道识别方法,以降低弯道处交通事故发生率。采用Hough变换求出车道线直线方程,根据高速公路的特点建立适当的道路模型,确定对应直线段上的最低点和最高点;分别比较直线段两侧白色像素点的个数判断出曲线道路的弯曲方向,从而进一步确定曲线段上的特征点,同时对车道线候选特征点进行滤波;分段拟合车道线的直线段和曲线段,实现车道线的二维重建。实验表明,能识别不同形状的车道线,具有很高的实时性和准确性,且在不同道路干扰下对连续型分道线和虚线型分道线的检测都有良好的识别效果。

Recognition Algorithm of Curl Road Based on Feature Point Extraction

Facing the fact that high traffic accident occurred on expressway curl road, a recognition algorithm of expressway curl road based on feature point extraction is studied to reduce the traffic ac- cident rate on those curves. Firstly, linear equations of lane lines are obtained by adopting Hough transform. A curl road model is established according to the characteristics of expressway, from which the lowest point and the highest point of the corresponding lane line can be obtained. Then the curve direction of bent lanes is judged out by comparing the number of white pixels on both sides of line segment, and the feature points in the curvilinear section are extracted. The candidate feature points should be filtered here. Finally, 2D reconstruction of the curl roads is realized via pieeewise fitting the linear and curvilinear section. Experiments show that this method not only has high real - time performance and accuracy in recognizing lanes of different shapes, but also has good effect on detecting both continuous lanes and dashed lanes with different interferences.