基于边界点分布特征的夜间道路检测算法研究

Lane Detection Algorithm for Night Applications Based on Distribution Features of Boundary Dots

提出一种在结构化道路情况下的夜间道路标志线检测算法。选择邻域均值滤波、Sobel算子和最大熵分割算法对道路图像进行预处理。结合道路灰度图像与道路边界图像,分析夜间道路边界点的分布特征,将道路边界点划分为到4个点集。针对虚假道路边界点,应用多方向搜索方法予以消除。选用2D的直线模型,采用改进的Hough变换从预处理后的图像中获取道路几何特征参数,最终检测出道路边界。试验结果表明本算法可靠、有效,户外试验中具有很好的鲁棒性。

This paper introduces a detection algorithm of a traveling vehicle on a structured road with lane markers at night. The algorithm adopts neighbor average filtering, Sobel operator and threshold segmentation of maximum entropy to preprocess the original image. Integrating gray level image and edge image, the paper analyzes the distribution fea- tures of lane boundary dots at night and sorts boundary dots into 4 sets. Multiple-direction searching method is presented to eliminate the false lane boundary dots. The paper uses modified Hough transformation algorithm to obtain the feature parameters of the lane edge. Experiment results show that the algorithm is reliable and effective, and is robust in outdoor experiments.