【摘要】 为了满足无人驾驶车在高速公路行驶的实时性和鲁棒性要求,提出了一种基于道路区域分割的车道线检测方法.该方法分道路区域分割和车道线检测2个阶段.在道路区域分割阶段,首先提取的道路颜色值,然后在二值边缘图像中搜索连通域,通过将连通域的颜色特征值与道路颜色特征值比较来快速定位道路区域,并将这一区域划定为车道线检测的感兴趣区域.车道线检测阶段则使用改进的概率Hough变换方法提取车道线点,并使用最小二乘法对车道线点集进行拟合,获得车道线模型的参数.实验证明该方法相比传统的利用标准Hough变换算法准确率提升23%,有效地排除了道路区域外的直线像素干扰,具备较好的鲁棒性和实时性.更多还原

【Abstract】 A lane detection method was developed based on road segmentation to satisfy the real-time and robustness requirements of an automated vehicle when driven on the highway.The proposed method was divided into two phases of road segmentation and lane detection.After extracting the color feature of the highway,the road segmentation phase quickly located the road region through searching connected regions of binary edge images and comparing the color feature of connected regions with the highway,and then set the road region as the region of interest(ROI) for lane detection.An improved probabilistic Hough transformation method was used to extract the lane pixels in the lane detection phase,and the least squares method was used for fitting the lane pixels to get the parameters of the lane model.Experiments show that the accuracy is improved 23% by the proposed method compared with the conventional method using standard Hough transformation,and linear pixels outside the road region were effectively ruled out.The proposed method has high robustness and real-time performance.更多还原