摘要
针对当前车道线检测算法在阴影及不均匀光照等复杂情况下的目标检测准确性低的问题,提出了一种边缘信息耦合改进的Hough变换的车道线实时检测方法。首先,利用Sobel算子计算车道图像的边缘梯度幅值,并利用2×2掩模测量其梯度方向,以提取车道线的边缘信息。根据得到的边缘信息,在搜索区域中计数每个行中的边缘像素的数目,将边缘像素最多的行作为感兴趣区域(RoI)的分界线,以确定RoI。然后,为了抑制非车道线边缘等杂乱背景的影响,选择特定的梯度方向进行细化RoI。最后,利用方向区间与阈值对Hough变换改进,将其应用于边缘像素,以提取车道线。并在Caltech数据集上进行了测试,数据表明,与当前流行车道线检测方案比较,所提方案在阴影、不均匀光照等不同道路情况下对车道线具有更高的检测精度与效率。
Aiming at the low accuracy of lane detection in complex situations such as shadows and uneven illumination in current lane detection algorithms,a real-time lane detection scheme based on edge information coupling and improved Hough transform was defined.Firstly,Sobel operator was used to calculate the amplitude of edge gradient,and 2×2 mask is used to measure the direction of gradient and extract the edge information of lane line.Secondly,according to the obtained edge information,the number of edge pixels in each row was counted in the line search area,and the boundaries of the region of interest(RoI)with the most edge pixels were selected to determine the RoI region.Then,in order to suppress the effect of cluttered background on the edges of non-lane lines,a specific gradient direction was selected to refine RoI.Finally,the direction interval and threshold were used to improve the Hough transform and to extract lane lines from edge pixels.Experiments show that compared with current popular lane detection schemes,the proposed scheme can achieve good lane detection under different road conditions such as shadows and uneven lighting.The algorithm can effectively improve the real-time and stability of lane departure warning system.
作者
付利军
兰方鹏
Fu Lijun;Lan Fangpeng(Department of Information Technology,Shanxi Yuncheng Agricultural Vocational and Technical College,Yuncheng044000,China;College of Computer Science and Technology,Taiyuan University of Technology,Taiyuan030600,China)
出处
《电子测量与仪器学报》
CSCD
北大核心
2019年第8期166-172,共7页
Journal of Electronic Measurement and Instrumentation
基金
山西省自然科学基金(2013011121-1)
山西省基础研究项目(2015021106)
山西省科技厅项目(2015021106)资助项目
