基于动态曲线的车道检测算法

Lane Detection Algorithm Based on Dynamic Curve

摘要: 针对无人驾驶技术中关于复杂车道线的检测问题，本文提出了一种基于车道线的颜色梯度变化和视频每帧图像的关联性的复杂车道线检测算法。该算法通过梯度检测的方法对原图像进行边缘像素提取，通过动态曲线的限制对所得的边缘像素进行特征点提取，筛选出目标边缘像素，去除非目标边缘像素。最后通过曲线拟合检测出车道线位置。通过现场实测表明，此算法最终能够通过曲线拟合得到较为准确的复杂车道线的位置信息。同时，实测结果也说明了此算法相比于其他算法而言具有更好的是实时性和鲁棒性。

Abstract: Aiming at the detection of complex lanes in unmanned driving technology, this paper proposes a complex lane detection algorithm based on the color gradient change of lanes and the correlation of each video frame. The algorithm extracts the edge pixels of the original image by gradient detection method, extracts the feature points of the edge pixels through the restriction of dynamic curve, filters out the target edge pixels and removes the non-target edge pixels. Finally, the lane location is detected by curve fitting. Field measurements show that the algorithm can finally obtain more accurate location information of complex lane lines through curve fitting. At the same time, the measured results also show that this algorithm has better real-time performance and robustness than other algorithms.