基于计算机视觉的车辆闯红灯行为检测研究
Research on Detection of Red Light Violation Based on Computer Vision
摘要:
为了实现对车辆闯红灯行为的自动智能和准确检测。本文提出了一种基于计算机视觉的车辆闯红灯行为检测方法。首先,采用高清摄像头实时采集道路上的车道、车辆和信号灯的视频图像;接着,对视频进行图像信息预处理,定制虚拟检测线、交通灯检测区域、车道检测区域,将定制结果保存在存储单元中;然后用计算机视觉技术来检测车辆闯红灯行为,通过HSV颜色空间中的H分量来识别信号灯的状态,利用MHI算法来识别车辆状态;最后,根据交通法规来识别车辆闯红灯行为。实验结果表明,本文提出的检测算法方法简单、识别率高、实现成本低并具有较高的鲁棒性。
Abstract: To automatically and accurately detect red-light violation, this paper presents a method of detection of red Light violation based on computer vision. Firstly, the image of traffic intersection is got by a monocular high-definition camera in real time. In order to detect the valid area, the image is pre-processed with the customization of stop line, traffic light areas, lane areas, and the result of customization is saved. Then, the state of traffic signals is recognized with H in HSV color model and the vehicles are detected by MHI. Finally, according to traffic laws and regulations, vehicle red-light violation will be identified. Experiment results show that this method has lightly computational load, low-cost, high recognition rate and robustness features.
参考文献
[1]
|
杨明. 面向智能交通系统的图像处理[J]. 计算机工程与应用, 2001, 9: 4-7.
|
[2]
|
夏劲, 郭红卫. 国内外城市智能交通系统的发展概况与趋势及启示[J]. 科技进步与对策, 2003, 1: 176-179.
|
[3]
|
周妮娜, 土敏, 黄心汉, 吕雪峰, 万国红. 车牌字符识别的预处理算法[J]. 计算机工程与应用, 2003, 15: 220-232.
|
[4]
|
张苗, 姚明海, 顾勤龙. 车辆牌照识别系统的一个新的实现方法[J]. 控制工程, 2003, 10(1): 59-61.
|
[5]
|
左奇, 史忠科. 一种基于数学形态学的实时车牌图像分割方法[J]. 中国图像图形学报, 2003, 8(3): 281-285.
|
[6]
|
张引. 基于空间分布最大类间方差的牌照图像二值化算法[J]. 浙江大学学报, 1997, 35(3): 272-276.
|
[7]
|
傅一平, 李志能, 袁丁. 基于 HIS 空间的颜色算法提牌照识别的性能[J]. 计算机工程与设计, 2004, 25(5): 703-707.
|
[8]
|
孟繁锋, 屈桢深, 曾庆双等. isMHI结合CAMSHIFT的多目标跟踪[J]. 计算机应用, 2009, 29(6): 1673-1676.
|
[9]
|
R. B. Gary, W. D. James. Motion seg-mentation and pose recog- nition with motion history gradients. Machine Vision and Appli- cation, 2002, 13(7): 174-184.
|
[10]
|
刘军学, 屈桢深, 任行行, 郭隽, 闻帆. 基于改进运动历史图像的多运动目标实时跟踪[J]. 计算机应用, 2008, 28(s1): 198- 201.
|