摘要
为提高车道线检测的准确性,根据车道图像信息前景区域纹理特征较为清晰的特点,提出了一种准确的车道线检测方法。针对车道中车辙印对应的纹理特征,通过不同尺度不同方向的Gabor模板变换分析,得到5个尺度、8个方向的特征图,并将其进行融合,得到像素点的纹理方向特征及其方向索引值;利用每一像素点的方向索引值对候选消失点进行投票,确定道路消失点,进而由消失点确定一个扇形扫描区域,定位车道线,建立其直线方程。实验表明该方法对强光照、阴影、车道线遮掩等情况有较强的鲁棒性。
In order to improve the accuracy of the lane line detection, an accurate lane line detection method was proposed, according to the feature that the texture features in the foreground region of lane image information were clearer. Aiming at the corresponding texture features in the lane left by rut, the characteristic diagrams of 5 scales and 8 directions were ob- tained by Gabor transform analysis with different scales at different directions. The characteristic diagrams were fused and texture direction characteristic together with index value of each pixel was obtained. The direction index value of each pixel was used to vote for the vanishing point, and then the vanishing point of the road was determined. Furthermore, a sector scanning area was determined by the vanishing point, and the lane marking was positioned. At last, the equation was established. The experimental results show that the method has strong robustness in strong light, shadow and lane line cloak.
出处
《重庆交通大学学报（自然科学版）》
CAS
北大核心
2014年第5期116-119,共4页
Journal of Chongqing Jiaotong University(Natural Science)
关键词
交通工程
车道线
GABOR变换
纹理特征
消失点
traffic engineering
lane marking
Gabor transform
texture feature
vanishing point
