刘景锋 李炎亮 江奎 郭雨婷



摘 要: 针对智能车辆在不同环境条件下识别道路车道线出现偏移的问题,提出了一种基于改进逆透视变换的车道线检测算法。首先利用HSL和Lab融合模型提取车道线颜色特征,经过二值化处理后利用透视变换将图像转换为鸟瞰图,然后根据二值图建立直方图,进行车道线位置的粗定位。最终通过滑动窗口算法以及直线拟合等处理,实现对车道线的精准识别。通过实验结果对比分析,提出的车道线检测算法能够解决车道线识别偏移问题。
关键词: 车道线检测; 逆透视变换; 滑动窗口; 特征融合
文章编号: 2095-2163(2021)03-0109-04 中图分类号:U461.99 文献标志码:A
【Abstract】Aiming at the problem that smart vehicles recognize the deviation of road lane lines under different environmental conditions, a lane line detection algorithm based on improved inverse perspective transformation is proposed. First, the HSL and Lab fusion model is used to extract the color features of the lane line, and after the binarization process, the image is converted into a bird's-eye view using perspective transformation. Then a histogram is established based on the binary image to perform coarse positioning of the lane line position. Finally, through sliding window algorithm, straight line fitting and other processing, the accurate recognition of lane lines is realized. Through comparative analysis of experimental results, the proposed lane line detection algorithm can solve the problem of lane line recognition offset.
【Key words】 lane line detection; inverse perspective transformation; sliding window; model fusion
0 引 言
為提高智能车驾驶安全性,实现自适应巡航、跟车行驶和车道保持等功能,实时精确的感知车道线信息是智能车自主安全行驶的基础。基于视觉算法是识别车道线常用的方法。
目前关于车道线识别的研究有很多。文献[1]提出了一种可以适应多路况的车道线识别方法,通过利用改进的TLC预警模型,提高了检测精度以及预警精度。刘萍等人[2]提出反透视变换算法解决道路遮挡干扰问题。彭浪[3]提出基于目标特征蒸馏的车道线检测算法,通过蒸馏技术和添加的解码器获取更准确的车道线信息。唐阳山等人[4]基于Roberts算子提高对车道线图像的精准分割。蒋一国[5]提出了基于GAN的车道线检测算法,利用pix2pix将车道线识别转化为图像的自动生成任务,然后提取细节特征。谢昌刚等人[6]提出了车道线隔行检测算法,通过动态确定ROI区域,识别效率提高40%,精度高达88.57%。……