直线模型下的高速公路车道线检测方法研究

Motorway Lane Detection Method Based on Straight-line Model

为了准确识别出高速公路图像中车道的标志边缘,结合道路图像特点,首先采用Sobel边缘算子对中值滤波后的道路图像进行增强,进而用Otsu算法将图像分割,在感兴趣区域运用基于像素的运算方法去除噪声,最后利用Hough变换进行拟合,识别出车道线。结果表明,对于2616×3488像元的道路图像,特征点与拟合直线的最大偏差为3.7263像元,能够准确识别出图像中的车道线,为智能汽车的安全预警提供了有益的探索。

In order to identify the edge in the motorway lane accurately, with the characteristics of road image, the road image which is filtered by median filter is enhanced by the Sobel operators. And then the image is segmented using Otsu algorithm. In the region of interest, the noise is removed using the computational method based on the pixels. At last, the lanes were identified by using Hough transform. The maximum offset between characteristic points and the fitting straight line is 3.7263 pixels. The experimental results are presented showing that the lanes can be identified accurately for 2616 × 3488 road image. Intelligent vehicle safety warning system is provided a useful exploration.