基于视频图像的车辆目标检测及速度测量
|
摘要
随着数字图像处理技术和计算机视觉技术的发展,视频检测技术逐渐成为智能交通系统中求取车辆速度等交通参数的重要方法,得到越来越多人的重视和肯定。准确交通信息的获取是智能交通系统及其子系统实现其各项功能的基础,所以利用视频检测技术来检测运动车辆目标、采集车速等实时交通参数成为当前智能交通系统研究与开发的重点之一。
本文根据从城市交通道路上采集的交通视频序列,参考视频检测方面现有的研究成果,提出两种运动目标检测方法,并通过对检测到的目标进行定位和跟踪,实现了车辆速度测量。论文主要工作包括:
1、本文针对智能交通系统中的交通视频序列,在现有的运动目标检测技术的基础上,提出了一种基于帧间差分法、图像阈值分割和数学形态学的运动目标检测方法。并对该方法的内容和实现过程进行了详细的说明。
2、本文研究了基于图像分割的目标分割方法。结合智能优化算法—人工免疫系统和遗传算法,本文提出一种改进的遗传算法—克隆选择遗传算法。该算法将克隆选择代替遗传算法中的概率选择,避免了改进后的遗传算法出现早熟收敛现象。仿真实验证明,基于克隆选择遗传算法的图像阈值分割对处理不同类型的图像具有良好的分割效果和较强的适应能力,也进一步验证了克隆选择遗传算法的有效性和实用性。
3、本文采用一种亚像素定位技术—灰度重心法对运动目标进行定位,并通过匹配连续两帧中同一车辆目标的最小外接矩形来实现目标跟踪。最后,利用不同帧的车辆自身长度的比例系数来得到车辆的行驶距离,计算车速。
4、在Matlab仿真实验环境下,建立了视频检测系统的实验平台,编程实现了运动目标检测过程,以及运动车辆速度的测量。仿真实验验证了本文所提方法的正确性和有效性。
With the development of digital image processing technology and computer vision technology, video detection technology becomes an important method to obtain traffic parameters such as vehicle speed in the Intelligent Transport System, and gains regard and affirmation by more and more people. To obtain exact traffic information is the foundation of carrying out all functions in the Intelligent Transport System and its subsystems, therefore, using video detection technology to detect moving vehicle objects、collect real-time traffic parameters such as vehicle speed becomes one of emphases in the research and exploiture of the Intelligent Transport System currently.
According to the traffic video sequence which collected from city traffic road, consulting study works in video detection, this paper proposed two kinds of moving objects detection methods, and implemented vehicle speed measurement after locating and tracking the detected objects. The major work implemented in this paper is presented as follows:
1、Based on current moving objects detection technology, this paper proposed a moving objects detection method based on difference in frames、image threshold segmentation and mathematical morphology, which is aimed at traffic video sequence in the Intelligent Transport System. Besides, the content and procession of this method are illuminated in detail.
2、This paper studied object segmentation method which is based on image segmentation. Combined with intelligent optimization algorithms—Artificial Immune System(AIS) and Genetic Algorithm(GA), this paper proposed an improved Genetic Algorithm called Clonal Selection Genetic Algorithm(CSGA). To avoid the improved Genetic Algorithm appearing premature convergence, this algorithm replaced probability selection in Genetic Algorithm by clonal selection. Emulational experimental results show that, image threshold segmentation based on CSGA has good segmentation effects and strong flexibility,as well as its validity and practicability.
3、This paper adopted a kind of subpixel location technologies—gravity method of the gray scale for locating moving objects, and through matching the smallest enclosing rectangle of same vehicle object in two consecutive frames to implement objects tracking. finally, vehicle length proportion coefficient in different frames was used to obtain vehicle moving distance and calculate speed.
4、Under the Matlab environment, experiment platform of video detection system was established, moving objects detection processing was implemented, as well as moving vehicle speed measurement. Emulational experimental results showed exactness and validity of the methods mentioned in this paper.
本文根据从城市交通道路上采集的交通视频序列,参考视频检测方面现有的研究成果,提出两种运动目标检测方法,并通过对检测到的目标进行定位和跟踪,实现了车辆速度测量。论文主要工作包括:
1、本文针对智能交通系统中的交通视频序列,在现有的运动目标检测技术的基础上,提出了一种基于帧间差分法、图像阈值分割和数学形态学的运动目标检测方法。并对该方法的内容和实现过程进行了详细的说明。
2、本文研究了基于图像分割的目标分割方法。结合智能优化算法—人工免疫系统和遗传算法,本文提出一种改进的遗传算法—克隆选择遗传算法。该算法将克隆选择代替遗传算法中的概率选择,避免了改进后的遗传算法出现早熟收敛现象。仿真实验证明,基于克隆选择遗传算法的图像阈值分割对处理不同类型的图像具有良好的分割效果和较强的适应能力,也进一步验证了克隆选择遗传算法的有效性和实用性。
3、本文采用一种亚像素定位技术—灰度重心法对运动目标进行定位,并通过匹配连续两帧中同一车辆目标的最小外接矩形来实现目标跟踪。最后,利用不同帧的车辆自身长度的比例系数来得到车辆的行驶距离,计算车速。
4、在Matlab仿真实验环境下,建立了视频检测系统的实验平台,编程实现了运动目标检测过程,以及运动车辆速度的测量。仿真实验验证了本文所提方法的正确性和有效性。
With the development of digital image processing technology and computer vision technology, video detection technology becomes an important method to obtain traffic parameters such as vehicle speed in the Intelligent Transport System, and gains regard and affirmation by more and more people. To obtain exact traffic information is the foundation of carrying out all functions in the Intelligent Transport System and its subsystems, therefore, using video detection technology to detect moving vehicle objects、collect real-time traffic parameters such as vehicle speed becomes one of emphases in the research and exploiture of the Intelligent Transport System currently.
According to the traffic video sequence which collected from city traffic road, consulting study works in video detection, this paper proposed two kinds of moving objects detection methods, and implemented vehicle speed measurement after locating and tracking the detected objects. The major work implemented in this paper is presented as follows:
1、Based on current moving objects detection technology, this paper proposed a moving objects detection method based on difference in frames、image threshold segmentation and mathematical morphology, which is aimed at traffic video sequence in the Intelligent Transport System. Besides, the content and procession of this method are illuminated in detail.
2、This paper studied object segmentation method which is based on image segmentation. Combined with intelligent optimization algorithms—Artificial Immune System(AIS) and Genetic Algorithm(GA), this paper proposed an improved Genetic Algorithm called Clonal Selection Genetic Algorithm(CSGA). To avoid the improved Genetic Algorithm appearing premature convergence, this algorithm replaced probability selection in Genetic Algorithm by clonal selection. Emulational experimental results show that, image threshold segmentation based on CSGA has good segmentation effects and strong flexibility,as well as its validity and practicability.
3、This paper adopted a kind of subpixel location technologies—gravity method of the gray scale for locating moving objects, and through matching the smallest enclosing rectangle of same vehicle object in two consecutive frames to implement objects tracking. finally, vehicle length proportion coefficient in different frames was used to obtain vehicle moving distance and calculate speed.
4、Under the Matlab environment, experiment platform of video detection system was established, moving objects detection processing was implemented, as well as moving vehicle speed measurement. Emulational experimental results showed exactness and validity of the methods mentioned in this paper.
引文
[1]李晖,贾辉然,牛晓辉,宋士普.智能交通系统的研究现状及发展趋势[J].产业与科技论坛. 2008,7(8):166
[2]鲍晓东.智能交通系统的现状及发展[J].北京工业职业技术学院学报.2007,6(2):42-45
[3]林仲扬.漫谈雷达测速[J].国外电子元器件.2006,10:71-73
[4]彭春华,刘建业,刘乐峰等.车辆检测传感器综述[J].传感器与微系统.2007,26(6):4-7
[5]孙玲玲,靖固,张科.视频测速技术的研究与仿真[J].科技信息.2009,1:511-512
[6]徐伟,王朔中.基于视频图像Harris角点检测的车辆测速[J].中国图像图形学报.2006,11(11):1650-1652
[7]李晴,徐群.复杂场景下多运动目标速度检测技术的实现[J].计算机与数字工程2006,34(11):167-171
[8] Barron J,Fleet D,Beauchemin S. Performance of optical flow techniques[J]. A International Journal of Computer Vision.1994,12(1):42-77
[9]卢宇,祝豪等.一种基于背景配准的光流场算法[J].光电技术应用. 2008,29(1):126-130
[10]计文平,郭宝龙等.基于虚拟线圈的光流法车流量检测[J].计算机仿真.2004,21(4):109-110
[11]于成忠,朱骏,袁晓辉.基于背景差法的运动目标检测[J].东南大学学报(自然科学版). 2005,35:159-161
[12]张红梅.基于背景差分的运动车辆检测的栅格算法研究[J].科学技术与工程.2008,8(20):5690-5693
[13]Yang Wei,Zhang Tianwen. A new method for the detection of moving targets in complex scenes[J].Computer Research & Development.1998,35(8):724-728.
[14]朱明旱,罗大庸等.帧间差分与背景差分相融合的运动目标检测算法[J].计算机测量与控制. 2005,13(3):215-217
[15]李惠松,王小铭等.一种基于帧间差分与时空相关性分析的运动目标检测算法[J].计算机与数字工程.2007,35(12):32-34
[16]唐佳林,李熙莹等.一种基于帧差法的夜间车辆检测方法[J].计算机测量与控制.2008,16(12):1811-1813
[17]章毓晋.图像分割[M].科学出版社,2001
[18]刘勍，应隽，周珑，温志贤.数学图像分割技术及其进展[J].天水师范学院学报. 2007,27(5):35-39
[19]付忠良.图像阈值选取方法-Otsu方法的推广[J].计算机应用.2000,20(5):37-39
[20]潘建平,邬明权.基于数学形态学的道路提取[J].计算机工程与应用.2008,44(11):232-233
[21]李弼程,彭天强,彭波等.智能图像处理技术[M].北京:电子工业出版社,2004
[22]丁永生.计算智能—理论、技术与应用[M].科学出版社.2004
[23]莫宏伟,左兴权,毕晓君.人工免疫系统研究进展[J]智能系统学报. 2009,4(1):21-29
[24]李敏强,寇纪淞等.遗传算法的基本理论与应用[M].科学出版社,2002
[25]保文星,朱长纯,崔万照.基于克隆选择的混合遗传算法在碳纳米管结构优化中的研究[J].物理学报. 2005,54(11):5281-5287
[26]张怀柱,向长波,宋建中,乔双.改进的遗传算法在实时图像分割中的应用[J].光学精密工程. 2008,16(2):333-337
[27]江中央,蔡自兴,王勇.用于全局优化的混合正交遗传算法[J].计算机工程.2009,35(4):204-209
[28]刘学增,周敏.改进的自适应遗传算法及其工程应用[J].同济大学学报(自然科学版).2009,37(3):303-307
[29]焦李成,杜海峰.人工免疫系统进展与展望[J].电子学报. 2003,31(10):1540-1548
[30]Leandro N de Castro,Fernando J Von Zuben. Learning and optimization using the clonal selection principle[J]. IEEE Trans on Evolutionary Computation. 2002, 6(3):239-251
[31]Leandro N de Castro,Fernando J Von Zuben. The Clonal Selection Algorithm with Engineering Applications[J]. Proc.of GECC’00, Workshop on Artificial Immune Systems and Their Applications. 2000:36-37
[32]Haifeng Du,Li-cheng Jiao,Maoguo Gong,and Ruochen Liu. Adaptive Dynamic Clone Selection Algorithms[C]. Springer-Verlag Berlin Heidelberg. 2004:768-773
[33]唐超礼,黄友锐,曲立国.自适应小生境克隆选择算法及其仿真研究[J].系统仿真学报. 2008,20(11):2956-2959
[34]尚荣华,焦李成,马文萍,张伟.求解多目标0/1背包问题的克隆选择算法[J].西安交通大学学报.2008,42(2):156-160
[35]唐守勤.基于数字图像处理精确测量的研究与实现[D].辽宁工学院.辽宁工学院信息科学与工程学院.2007:51-54
[36]刘萌萌.基于无标度摄像机的车流跟踪与速度估计算法[J].计算机技术与发展.2008,18(6):111-114
[2]鲍晓东.智能交通系统的现状及发展[J].北京工业职业技术学院学报.2007,6(2):42-45
[3]林仲扬.漫谈雷达测速[J].国外电子元器件.2006,10:71-73
[4]彭春华,刘建业,刘乐峰等.车辆检测传感器综述[J].传感器与微系统.2007,26(6):4-7
[5]孙玲玲,靖固,张科.视频测速技术的研究与仿真[J].科技信息.2009,1:511-512
[6]徐伟,王朔中.基于视频图像Harris角点检测的车辆测速[J].中国图像图形学报.2006,11(11):1650-1652
[7]李晴,徐群.复杂场景下多运动目标速度检测技术的实现[J].计算机与数字工程2006,34(11):167-171
[8] Barron J,Fleet D,Beauchemin S. Performance of optical flow techniques[J]. A International Journal of Computer Vision.1994,12(1):42-77
[9]卢宇,祝豪等.一种基于背景配准的光流场算法[J].光电技术应用. 2008,29(1):126-130
[10]计文平,郭宝龙等.基于虚拟线圈的光流法车流量检测[J].计算机仿真.2004,21(4):109-110
[11]于成忠,朱骏,袁晓辉.基于背景差法的运动目标检测[J].东南大学学报(自然科学版). 2005,35:159-161
[12]张红梅.基于背景差分的运动车辆检测的栅格算法研究[J].科学技术与工程.2008,8(20):5690-5693
[13]Yang Wei,Zhang Tianwen. A new method for the detection of moving targets in complex scenes[J].Computer Research & Development.1998,35(8):724-728.
[14]朱明旱,罗大庸等.帧间差分与背景差分相融合的运动目标检测算法[J].计算机测量与控制. 2005,13(3):215-217
[15]李惠松,王小铭等.一种基于帧间差分与时空相关性分析的运动目标检测算法[J].计算机与数字工程.2007,35(12):32-34
[16]唐佳林,李熙莹等.一种基于帧差法的夜间车辆检测方法[J].计算机测量与控制.2008,16(12):1811-1813
[17]章毓晋.图像分割[M].科学出版社,2001
[18]刘勍，应隽，周珑，温志贤.数学图像分割技术及其进展[J].天水师范学院学报. 2007,27(5):35-39
[19]付忠良.图像阈值选取方法-Otsu方法的推广[J].计算机应用.2000,20(5):37-39
[20]潘建平,邬明权.基于数学形态学的道路提取[J].计算机工程与应用.2008,44(11):232-233
[21]李弼程,彭天强,彭波等.智能图像处理技术[M].北京:电子工业出版社,2004
[22]丁永生.计算智能—理论、技术与应用[M].科学出版社.2004
[23]莫宏伟,左兴权,毕晓君.人工免疫系统研究进展[J]智能系统学报. 2009,4(1):21-29
[24]李敏强,寇纪淞等.遗传算法的基本理论与应用[M].科学出版社,2002
[25]保文星,朱长纯,崔万照.基于克隆选择的混合遗传算法在碳纳米管结构优化中的研究[J].物理学报. 2005,54(11):5281-5287
[26]张怀柱,向长波,宋建中,乔双.改进的遗传算法在实时图像分割中的应用[J].光学精密工程. 2008,16(2):333-337
[27]江中央,蔡自兴,王勇.用于全局优化的混合正交遗传算法[J].计算机工程.2009,35(4):204-209
[28]刘学增,周敏.改进的自适应遗传算法及其工程应用[J].同济大学学报(自然科学版).2009,37(3):303-307
[29]焦李成,杜海峰.人工免疫系统进展与展望[J].电子学报. 2003,31(10):1540-1548
[30]Leandro N de Castro,Fernando J Von Zuben. Learning and optimization using the clonal selection principle[J]. IEEE Trans on Evolutionary Computation. 2002, 6(3):239-251
[31]Leandro N de Castro,Fernando J Von Zuben. The Clonal Selection Algorithm with Engineering Applications[J]. Proc.of GECC’00, Workshop on Artificial Immune Systems and Their Applications. 2000:36-37
[32]Haifeng Du,Li-cheng Jiao,Maoguo Gong,and Ruochen Liu. Adaptive Dynamic Clone Selection Algorithms[C]. Springer-Verlag Berlin Heidelberg. 2004:768-773
[33]唐超礼,黄友锐,曲立国.自适应小生境克隆选择算法及其仿真研究[J].系统仿真学报. 2008,20(11):2956-2959
[34]尚荣华,焦李成,马文萍,张伟.求解多目标0/1背包问题的克隆选择算法[J].西安交通大学学报.2008,42(2):156-160
[35]唐守勤.基于数字图像处理精确测量的研究与实现[D].辽宁工学院.辽宁工学院信息科学与工程学院.2007:51-54
[36]刘萌萌.基于无标度摄像机的车流跟踪与速度估计算法[J].计算机技术与发展.2008,18(6):111-114