为快速建立道路精确的几何模型,利用车载移动测量系统获取的序列影像开展了道路边线自动提取方法研究。针对不同类型道路边线的特征,提出了一种基于序列影像种子跟踪的道路线自动算法。首先,对不同质量道路影像分别采用基于灰度信息和梯度信息的道路边线边缘检测方法;其次,采用线性回归模型拟合道路边线,以当前帧拟合直线为种子点,提出了基于种子直线的序列影像感兴趣预测方法,在此基础上对下一帧影像道路边线进行跟踪;最后,以高速公路和城市街区道路序列影像为例,开展了道路边线自动跟踪提取试验,结果表明此方法可以达到优于20cm的精度,具有很好的鲁棒性。
In order to quickly build precise road geometric model, an automatic extraction approach of road boundary line by using mobile mapping system is studied. A road line automatic extraction algorithm using seed tracking with sequence image for different road boundary line features is proposed. First, a road boundary line edge detection method based on gray information and gradient information for different quality images is used. Second, a linear regression model is used to fit the road boundary line. Taking the fitted road line in current image frame as the line seeds, a prediction method of the region of interest (ROI) of the sequence image frames is presented to track the road boundary line in the next image frame. Finally, the road boundary line automatic tracking and extraction experiment is conducted with the proposed algorithm on the selected expressway and street road sequence images. The result shows that the proposed method can achieve the accuracy better than 20 cm with good robustness.