欢迎访问《汽车安全与节能学报》，
汽车安全与节能学报

JASE ›› 2019, Vol. 10 ›› Issue (4): 459-466.DOI: 10.3969/j.issn.1674-8484.2019.04.007

• 汽车安全 • 上一篇    下一篇

用于智能汽车的复杂光照环境车道线检测及跟踪方法

金智林1，何麟煊1，赵万忠1, 2#br#   

  1. （1. 南京航空航天大学 能源与动力学院，南京 210016，中国；
     2. 清华大学 汽车安全与节能国家重点实验室，北京 100084，中国）
  • 收稿日期:2019-04-15 出版日期:2019-12-31 发布日期:2020-01-01
  • 作者简介:第一作者 / First author ： 金智林 (1978 －)，男( 汉)，江西，副教授。E-mail: jinzhilin@nuaa.edu.cn。
  • 基金资助:
    国家自然科学基金资助项目（ 51775269）；汽车安全与节能国家重点实验室开放基金（KF1812）。

Detection and tracking method of lane line for intelligent vehicles under complex illumination condition

JIN Zhilin 1, HE linxuan 1, ZHAO Wanzhong 1, 2   

  • Received:2019-04-15 Online:2019-12-31 Published:2020-01-01

PDF

278

摘要： 针对复杂光照环境的智能汽车行驶安全问题，提出一种车道线检测及跟踪的改进方法。在图 像预处理阶段，应用灰度变换方法增大不同光照环境的车道线和道路对比度；用改进概率 Hough变 换方法提取二值化图像中的车道线；用最小二乘法进行车道线拟合；根据前一帧车道线检测结果建立 Kalman 滤波动态感兴趣区域，实现车道线准确跟踪。进行夜间光照、弱光照、强光照及正常光照等 不同光照环境下的车道线检测和跟踪实验。结果表明：该方法的准确率约为 97.53%，对于单帧车道 图像的处理时间约为 60 ms，具有较好的准确性和实时性，并具有抗路面阴影、交通标志、车辆遮挡、 路灯等因素干扰的能力。

关键词: 智能汽车, 行驶安全, 车道线检测及跟踪, 复杂光照环境, 灰度变换, 改进概率 Hough变换,  Kalman 滤波

Abstract: An improved lane line detection and tracking method was proposed for the driving safety of intelligent vehicles under complex illuminations. Gray-level transformation was used for image preprocessing to increase the contrast between lane lines and roads. Lane lines in binary images were extracted by using a progressive probabilistic Hough transform, and were fitted by using a least square method. Dynamic region of interest was established based on Kalman-filter according to the results of the previous image to track the lane lines. The lane line detection and tracking experiments were carried out under different illumination conditions, such as night light, weak light, normal light and strong light. The results show that this method has an accuracy of 97.53% and a real-time performance of 60 ms in the detection and tracking of lane lines, and has robustness against the interference of road shadow, traffic signs, vehicles and other factors.

Key words: intelligent vehicles, driving safety, lane line detection and tracking, complex illumination, gray-level transformation, progressive probabilistic Hough transform, Kalman-filter