Vehicle Headlights Detection Using Markov Random Fields
Vehicle Headlights Detection Using Markov Random Fields
复制标题
DOI：
10.1007/978-3-642-12307-8_16
复制
发表时间：
2009-09
期刊：
影响因子：
--
通讯作者：
Wei Zhang;Q. M. J. Wu;Guanghui Wang
中科院分区：
文献类型：
--
作者：
Wei Zhang;Q. M. J. Wu;Guanghui Wang
基于视觉的交通监控是计算机视觉领域的一个重要课题。在夜间环境中，移动的车辆通常通过前灯检测。然而，强大的前大灯检测被路面上的强烈反射所阻碍。在本文中，我们提出了一种新的汽车前灯检测方法。首先，在分析邻域光衰减模型的基础上引入反射强度图；其次，利用拉普拉斯高斯滤波得到反射抑制映射。第三，将灰度强度、反射强度映射和反射抑制映射合并到马尔可夫随机场框架中，并使用迭代条件模式算法对该框架进行优化。在典型场景下的实验结果表明，该方法可以在强反射情况下正确检测出前照灯。定量评价表明，该方法优于现有方法。
Vision-based traffic surveillance is an important topic in computer vision. In the night environment, the moving vehicles are commonly detected by their headlights. However, robust headlights detection is obstructed by the strong reflections on the road surface. In this paper, we propose a novel approach for vehicle headlights detection. Firstly, we introduce aReflection Intensity Mapbased on the analysis of light attenuation model in neighboring region. Secondly, aReflection Suppressed Mapis obtained by using Laplacian of Gaussian filter. Thirdly, the headlights are detected by incorporating the gray-scale intensity,Reflection Intensity Map, andReflection Suppressed Mapinto a Markov random fields framework, which is optimized using Iterated Conditional Modes algorithm. Experimental results on typical scenes show that the proposed method can detect the headlights correctly in the presence of strong reflections. Quantitative evaluations demonstrate that the proposed method outperforms the existing methods.