期刊论文

Research of lane detection and tracking methods based on multi-sensor fusion

中文

计算机应用研究

1001-3695

2018

35

2

600-603,607

10.3969/j.issn.1001-3695.2018.02.059

国家自然科学基金资助项目（61573261） 2016年湖北省教学研究资助项目（201631）

本校为第一机构

车道线的有效检测与跟踪是智能车正确识别道路的前提。针对现有车道线检测与跟踪算法效率不高的难题,提出了一种基于视觉传感器与车道级高精度地图相融合的车道线检测与跟踪方法。该方法首先用改进的Hough变换提取边缘线段;然后基于滤波预测与更新车道线模型状态参数;最后结合高精度地图中车道线先验模型参数,跟踪车道线轨迹。现场实测结果表明,算法的实时性和鲁棒性满足算法性能评价体系的各项指标,较符合智能车对车道线检测的要求。

The precondition of the intelligent vehicles correctly identify road is lane of effective detection and tracking. In or- der to improve the performance of present lane line detection and tracking algorithm, this paper proposed a new method com- bing the visual sensor and lane-level high precision map. Firstly, it used the modified Hough transform to extract edges. Then, it estimated and updated the parameters of lane model based on filtering. Finally, it traced the trajectory of lanes through combing the priori model parameters of the lane-leve...