本文中提出了一种基于车道线特征的三车道检测算法。首先，在车道线预提取过程中对道路消失线以下部分的整个车道图像进行模糊化和边缘检测，并根据边缘点位置和方向角对消失点进行定位，同时基于消失点位置提取直线并结合车道模型对构成三车道的直线进行筛选和补充。接着在车道跟踪阶段，根据前一帧图像检测出的直线和消失点位置，对车道图像局部区域分别进行边缘点、直线的跟踪检测，并对消失点位置进行重定位。最后，对车道参数进行寻优以计算车道线曲率和车道宽度。试验结果表明，提出的边缘检测算法能有效检测模糊车道线边缘并抑制噪声，消失点和直线的检测方法耗时少且准确性高。在直线检测的基础上进行车道模型匹配能提高车道识别实时性，算法在车道线模糊、雨天、大雾和大曲率等环境下均具有较好的适应性。%A three-lane detection algorithm is proposed in this paper based on the features of lane mark-ings. Firstly in the process of lane markings pre-extraction, the blurrization and edge detection are conducted on the whole lanes image below road vanishing line, the position of vanishing points are determined according to the loca-tion and directional angle of edge points, based on which straight lines are extracted, and combined with lane model the straight lines composing three lanes are selected and added. Then in the stage of lane tracking, the edge points and straight lines in the local areas of lane image are tracked and detected according to the position of straight lines and vanishing points detected from previous frame of image, and the positions of vanishing points are updated. Fi-nally the lane parameters are optimized and the curvature of lane markings and lane width are calculated. The test results show that the edge detection algorithm proposed can effectively detect the blurred edges of lane markings with noise suppressed, and the vanishing points and straight lines detection method has the merits of less time-consuming and high accuracy. Lane model matching on the basis of straight lines detection can improve the real-time perform-ance of line detection, and under adverse environments such as blurred lane markings, rain, heavy fog and large curvature, the algorithm still has better adaptability.
展开▼
