基于多项式曲线拟合的车道线检测

Lane Line Detection Based on Perspective Transformation and Curve Fitting

针对无人驾驶行车车道线检测功能模块的实现问题,提出一种基于多项式曲线拟合算法的车道线检测方法。首先,对图像颜色空间进行变换,通过梯度预处理得到一个能基本消除背景噪声的二值图像;然后,经透视变换将二值图像处理成鸟瞰图像,通过检测鸟瞰图像的像素直方图曲线得到车道线的基点;最后使用纵向滑动窗口技术和多项式曲线拟合车道线方程来绘制车道线,将计算出的车道线像素点坐标映射到原图像中,并对高亮显示的车道线、车辆行驶区域、车道线曲率和车辆位置进行评估和视频展示。由对比实验可知,该方法能够解决无人驾驶中对车道线的检测和识别问题,且效果优于霍夫变换检测方法。

Aiming at the realization of the lane detection function module of unmanned driving lanes,this paper proposes a polynomial curve fitting algorithm based on perspective transformation and window sliding for lane detection.First,using a combination of gradient preprocessing such as color space transformation and gradient calculation to obtain a binary image that basically eliminates background noise,and then using perspective transformation to process the binary image into a bird s eye view.By detecting the bird s eye picture pixel histogram curve and get the base point of the lane line,and finally using the longitudinal sliding window technique and polynomial curve fitting lane line equation to draw the lane line curve,map the calculated lane line pixels back to the original image,and draw to the highlighted lane in the original image lines,vehicle travel areas,lane line curvatures and vehicle position assessments are shown in the original video.Through comparative experiments,this method can solve the problem of detection and recognition of lane lines in unmanned driving,and the effect is better than the algorithm of lane line detection by Hough transform technology.