基于离群值过滤道路中心线检测算法的研究

Research on road centerline detection algorithm based on outlier filtering

在实际应用中,针对智能驾驶对道路辅引线实时性与准确性的高要求,提出了一种基于离群值过滤的道路中心线提取算法。在图像预处理方面,将梯度阈值与颜色阈值相结合,降低在道路边缘缺陷或光照不足的影响;在边缘拟合方面,将滑窗检测与最小二乘法结合拟合道路边缘;在中心线提取方面,通过计算像素坐标的差值来达到离群值过滤的目的,经验证:算法在实时性与准确性方面都有一定的优势。

In practical application,aiming at the high requirements of real-time and accuracy of road auxiliary lead for intelligent driving,a road centerline extraction algorithm based on outlier filtering is proposed.In image preprocessing,gradient threshold and color threshold are combined to reduce the influence of road edge defects or insufficient illumination.In terms of edge fitting,under the inverse perspective transform image,the sliding window detection is combined with the least square method to fit the road edge.In the aspect of centerline extraction,outlier filtering is achieved by calculating the difference of pixelcoordinates.It is proved that the algorithm in this paper has certain advantagesin real-time and accuracy.