一种基于直线模型的车道线识别算法研究

Linear model based lane mark identification algorithm

为了更好地满足车道标志线识别算法的实时性和鲁棒性要求,提出了一种新的、有效的车道标志线识别算法。将图像灰度化后,采用中值滤波去除图像采集过程中引入的噪声,应用方向可调滤波器进行边缘提取,在提取过程中对原图像进行感兴趣区域划分并采用边缘分布函数法确定方向可调滤波器的初始方向角。提出使用基于梯度加权的霍夫变换对车道标志线进行识别,通过建立梯形感兴趣区域的方法实现对车道标志线的实时跟踪,并对多段实地采集的视频进行实验测试。结果表明:基于方向可调滤波器与梯度加权的霍夫变换相结合的车道标志线识别方法,简化了对车道标志线信息特征参量的估计;不仅大大缩减了算法的执行时间,而且使算法的鲁棒性得到很大的提高。

In order to meet the requirements of the real-time and robustness of lane mark identification algorithm,this paper proposed a lane mark identification algorithm.Turning the color image into grayscale,filtering out noise by median filter and extracting the lane mark edge by steerable filter were introduced firstly.To determine the initial direction angle of the filter,the image was divided into interest regions and the algorithm of edge distribution function（EDF） was used.Then,put forward the identification algorithm based on gradient weighted Hough transform.Finally,performed the real-time tracking of the lane mark by establishing trapezoid areas of interest,and verified the validity of the proposed method by experiments by using several videos.The results show that the combination of steerable filter and gradient weighted Hough transform simplifies the information characteristics of the lane mark estimation,which also greatly reduces the execution time and makes the robustness of the algorithm greatly improved.