基于改进SIS算法和LSD的车道线检测方法研究

为了降低车道线识别算法在车道线存在阴影遮挡、路面出现泛白现象等不同道路环境下的误检率,提出了一种基于改进简单图像统计(SIS)阈值算法和直线段检测(LSD)的车道线检测算法。首先,在图像预处理阶段采用改进的SIS阈值算法进行二值化。然后采用直线段检测(LSD)算法检测直线,通过平行线对来估计消失点位置并利用消失点去除干扰。最后,利用车道线连续性和车道间距确定车道线感兴趣区并精确确定车道线位置。分别采用加州理工学院的车道数据集和实际采集的城市道路、高速公路的视频对所提出方法进行了实验验证,实验结果表明,该算法误检率低,鲁棒性高,能在复杂环境下快速、准确识别车道线。

Lane Line Detection Method Research Based on Improved Algorithm of SIS and LSD

In order to reduce the false positive rate of the lane recognition algorithm in different road circumstances such as the lane is prone to the occlusion of shadow and the pavements appear blanched. A lane detection algorithm based on improved SIS threshold algorithm and line segment detection (LSD) is proposed. First of all, the improved SIS threshold algorithm is used to transform the image data into binary data, during the image preprocessing stage. Then the line segment detection (LSD) algorithm is applied to detect whether there exists straight line in the image. Moreover, the parallel lines are used to estimate the position of the vanishing point, which in turn eliminate the interference. Finally, the location of the lane line is determined in the lane line region of interest accurately through the analysis of the continuity of lane line and the values of lane spacing. The method is verified by using California institute of technology"s driveway data set and the actual city road and video of the highway. The results show that the algorithm has low false positive rate and high robustness. Moreover, the lane line can be identified precisely and quickly in the complex environment.