一种基于视觉的车道线检测及追踪方法

A Method of Lane Line Detection and Tracking Based on Vision

文章以OpenCV软件为主要平台,基于机器视觉建立一种车道线的检测及跟踪方法。首先检测出图片中车道线区域,对每一帧图片做透视转换到鸟瞰图视角。然后对鸟瞰图二值化,进一步区分左右两条车道线。用滑动窗口的方法检测第一帧的车道线像素点,从第一帧的曲线周围寻找后面的车道线像素点,拟合车道线曲线。经实验证明该方法能够准确地检测出车道线。

This paper uses OpenCV software as the main platform,and establishes a method of lane line detection and tracking based on machine vision.First,detect the lane line area in the picture,and convert each frame of the picture to the bird's-eye view perspective.Then binarize the bird's-eye view to further distinguish the left and right lanes.The sliding window method is used to detect the lane line pixels of the first frame,and find the subsequent lane line pixels around the curve of the first frame to fit the lane line curve.Experiments show that this method can accurately detect lane lines.