基于直线模型的车道线实时检测方法

摘要

To increase the accuracy and timeliness of lane detection, a linear model-based real-time lane detection approach is presented. Firstly, improved Sobel operator is used to obtain the edge characteristic points from the given road image, the adap-tive double threshold method is introduced to get the binary image; then the classic lane detection based on Hough transform (HT) is improved to obtain a better identification effect. Finally the size and location of the small window in region of interest (ROD are determined dynamically by adopting Kalman filter, which can guide lane track of the following frames. This approach is tested and the experimental results show that the proposed scheme is accuracy robust and fast enough for real time requirements.%为提高车道线检测的精度和实时性,提出了一种基于直线模型的实时车道线检测方法.采用改进的Sobel算子进行边缘检测,利用自适应双阈值的方法进行图像的二值化.对基于Hough变换的车道线检测方法进行了改进,令识别视觉效果大大提升.并利用Kalman滤波器来动态确定感兴趣小窗口的大小和位置,实现后续帧道路图像的跟踪.实验结果表明,该方法准确性高,具有较好的实时性和鲁棒性.