基于视频的静态背景下运动目标检测与跟踪研究
|
摘要
本文以基于静态背景下,对运动的物体进行识别、检测和跟踪,并鉴于软件编程的可实现性,在对图像进行预处理前提下,采用自适应高斯背景模型为静止背景下的图像序列提供背景图像,同时为了能更好的响应实际背景发生变化的情形,特别是对自适应背景,本文提出了算法上的改进和简化。
在图像预处理方面,主要分析了图像预处理处理方法,包括:中值滤波、维纳滤波、灰度处理、直方图均衡化技术,并对其进行了仿真。
在背景建立方面,通过对高斯背景模型的分析,对高斯模型加以改进,主要针对其方差和均值进行修改,变换样本均值和方差的计算方式,对其动态抽取样本,缩短取样时间和难度,尽快完成背景的初始化。
在目标检测方面,给出了运动目标检测与分割的仿真结果。在介绍了三种传统的运动目标检测算法的基本原理的基础上,运用帧间差分的方法,对目标进行提取,并对提取后的图像作出形态学处理,从而得到理想的目标提取图像。
在运动目标跟踪研究方面,本文从理论上分析卡尔曼滤波器的工作过程,基于人体运动特点和卡尔曼滤波的预测特性,提出一种基于人体中心的改进算法,提高预测的准确性,从而缩小搜索范围,减少搜索时间。
In this paper, based on a static background, the movement of objects to identify, detect and track, and in view of the software can be programmed to achieve, in the image of the pre-premise, adaptive Gaussian model for the background static background image Sequence to provide the background image, at the same time in order to be able to better respond to changes.
In the pre-image, the main image analysis of the pre-treatment, including: median filter, Wiener filter, to deal with gray, histogram equalization technology, and simulation.
In the context of the establishment, through the background of the Gaussian model analysis of the Gaussian model to improve, mainly for its mean and variance modified transform the sample mean and variance calculation, the dynamic of their sample to shorten the time sampling and the degree of difficulty as soon as possible to complete the background initialization.
In target detection, given the moving target detection and segmentation of the simulation results. After the introduction of the three traditional moving target detection algorithm based on the basic principle, the use of frame-difference approach to the extraction of targets, and after extracting the images to deal with morphology and thus be the ideal goal of extracting images.
Moving target in the follow-up study, this paper analyzed the work of the Kalman filter, based on the characteristics of human motion and Kalman filtering properties of the forecast, a human-based center for the Improvement of algorithms to improve the accuracy of the forecasts in order to reduce the scope of the search, the target of a search match.
在图像预处理方面,主要分析了图像预处理处理方法,包括:中值滤波、维纳滤波、灰度处理、直方图均衡化技术,并对其进行了仿真。
在背景建立方面,通过对高斯背景模型的分析,对高斯模型加以改进,主要针对其方差和均值进行修改,变换样本均值和方差的计算方式,对其动态抽取样本,缩短取样时间和难度,尽快完成背景的初始化。
在目标检测方面,给出了运动目标检测与分割的仿真结果。在介绍了三种传统的运动目标检测算法的基本原理的基础上,运用帧间差分的方法,对目标进行提取,并对提取后的图像作出形态学处理,从而得到理想的目标提取图像。
在运动目标跟踪研究方面,本文从理论上分析卡尔曼滤波器的工作过程,基于人体运动特点和卡尔曼滤波的预测特性,提出一种基于人体中心的改进算法,提高预测的准确性,从而缩小搜索范围,减少搜索时间。
In this paper, based on a static background, the movement of objects to identify, detect and track, and in view of the software can be programmed to achieve, in the image of the pre-premise, adaptive Gaussian model for the background static background image Sequence to provide the background image, at the same time in order to be able to better respond to changes.
In the pre-image, the main image analysis of the pre-treatment, including: median filter, Wiener filter, to deal with gray, histogram equalization technology, and simulation.
In the context of the establishment, through the background of the Gaussian model analysis of the Gaussian model to improve, mainly for its mean and variance modified transform the sample mean and variance calculation, the dynamic of their sample to shorten the time sampling and the degree of difficulty as soon as possible to complete the background initialization.
In target detection, given the moving target detection and segmentation of the simulation results. After the introduction of the three traditional moving target detection algorithm based on the basic principle, the use of frame-difference approach to the extraction of targets, and after extracting the images to deal with morphology and thus be the ideal goal of extracting images.
Moving target in the follow-up study, this paper analyzed the work of the Kalman filter, based on the characteristics of human motion and Kalman filtering properties of the forecast, a human-based center for the Improvement of algorithms to improve the accuracy of the forecasts in order to reduce the scope of the search, the target of a search match.
引文
[1]孟凤,王成儒,多模态背景下快速运动目标检测的研究[J],电子测量技术,2007,30(6):33一35
[2]孙即祥,图像分析[M],北京,科学出版社,2005
[3]阮秋琦等译,(美)冈萨雷斯等著,数字图像处理(第二版)[M],北京,电子出版社,2007
[4]周军妮,曹剑中,刘波等,一种基于局部最小代价分水岭变换的图像分割新方法[J],光子学报,2005,34(l):142一145
[5]周锐,复杂背景中的运动目标提取与跟踪[J],系统工程与电子技术,1997,9:10一13
[6]唐宜清,彭可,罗湘运,视频监控中一种完整提取运动目标的检测算法[J],计算技术与自动化,2007,26(2):42-45
[7]朱矿军,视频车辆检测技术在智能交通系统的应用,大连理工大学学报,2006
[8]何斌,马大予,王运坚,朱红莲,Visual C++数字图像处理,人民邮电出版社.
[9]秦永元等.卡尔曼滤波与组合导航原理.西北工业大学出版社,1998.11
[10]任臣,张覃平,基于Kalman滤波理论的运动目标检测新方法[J].光电工程. 2007, 34(4):7-11
[11]张旭东,钱玮,等.视频图像中运动目标的实时检测[J].系统工程与电子技术.2005,27(3):419-421
[12]周兵,李晓强,苏士美.实用户外视频监控异常检测策略[J].计算机工程.2003,29(16):16-18
[13]单勇.复杂条件下视频运动目标检测和跟踪[D].博士学位论文,国防科技大学,长沙, 2006
[14]孙吉花,刘肖琳.光照变化的背景减除新算法[J].计算机仿真..2007,24(9):214-217
[15]王栓,艾海周,何克忠.基于差分图象的多运动目标的检测与跟踪[J].中国图象图形学报,1999,1.4(6):24一29
[16]章毓晋.图像工程(下册)一图像理解与计算机视觉[M].北京:清华大学出版社2000年8月.
[17]郑南宁.计算机视觉与模式识别[M],北京:国防工业出版社,2006
[18]杨淑莹,VC++图像处理程序设计[M].北京,清华大学出版社,北京交通大学出版社,2005.1.
[19]求实科技.Visual C++一数字图像处理典型算法及实现[M].人民邮电出版社,2006.6.
[20]李刚,邱尚斌等.基于背景差法和帧间差法的运动目标检测方法[J].仪器仪表学报2006,27(8):64一69.
[21]韩思奇,王蕾.图像分割的闭值法综述[J].系统工程与电子技术.2000,6(24)91一102.
[22]赵荣祷编著.数字图像处理导论.西安:西北工业人学出版社,2000
[23]崔屹.《图像处理与分析一数学形态学方法及应用》.科学出版社,2000
[24] Lipton,and H.Fujiyoshi, etc. Moving target classification and Tracking from real-time video,Proc. IEEE Workshop Applications of Computer Vision,Oct 1998,8(14)
[25] Foresti,G L. Object recognition and tracking for remote video surveillance,IEEE transactions on Circuits and Systems for Video Technology, 1999,9(7):1045一1062
[26] P.Fieguth and D.Terzopoulos,”Color一based Tracking。Heads and other Mobile objects at video frame rates,”In Pore .Of IEEECVPR,pp.21一27,1997.
[27] Ko11er D.Weber J.Towards robust automatic traffic scene analysis in real-time [J」.Proceedings of the IEEE Workshop on Applications of Computer
[28] Zhou S K,Chellappa R, Moghaddan B.Visual tracking and recognition using appearance-adaptive models in particle filters. IEEE Transactions on Image Processing, 2004.
[29]陈朝阳,张桂林.基于图像对称差分运算的运动小目标检测方法[J].华中理工大学学报1998,26(9):34一35
[30]王新余,张桂林.基于光流的运动目标实时检测方法研究[J],计算机工程与应用,2004,l:43一46
[31]冯伟,视频序列运动目标检测与识别方法研究,硕士学位论文,西安,西北工业大学,2003
[32]谭优,王泽勇,图像闭值分割算法实用技术研究与比较[J],微计算机信息,2007,23(24):298一233
[2]孙即祥,图像分析[M],北京,科学出版社,2005
[3]阮秋琦等译,(美)冈萨雷斯等著,数字图像处理(第二版)[M],北京,电子出版社,2007
[4]周军妮,曹剑中,刘波等,一种基于局部最小代价分水岭变换的图像分割新方法[J],光子学报,2005,34(l):142一145
[5]周锐,复杂背景中的运动目标提取与跟踪[J],系统工程与电子技术,1997,9:10一13
[6]唐宜清,彭可,罗湘运,视频监控中一种完整提取运动目标的检测算法[J],计算技术与自动化,2007,26(2):42-45
[7]朱矿军,视频车辆检测技术在智能交通系统的应用,大连理工大学学报,2006
[8]何斌,马大予,王运坚,朱红莲,Visual C++数字图像处理,人民邮电出版社.
[9]秦永元等.卡尔曼滤波与组合导航原理.西北工业大学出版社,1998.11
[10]任臣,张覃平,基于Kalman滤波理论的运动目标检测新方法[J].光电工程. 2007, 34(4):7-11
[11]张旭东,钱玮,等.视频图像中运动目标的实时检测[J].系统工程与电子技术.2005,27(3):419-421
[12]周兵,李晓强,苏士美.实用户外视频监控异常检测策略[J].计算机工程.2003,29(16):16-18
[13]单勇.复杂条件下视频运动目标检测和跟踪[D].博士学位论文,国防科技大学,长沙, 2006
[14]孙吉花,刘肖琳.光照变化的背景减除新算法[J].计算机仿真..2007,24(9):214-217
[15]王栓,艾海周,何克忠.基于差分图象的多运动目标的检测与跟踪[J].中国图象图形学报,1999,1.4(6):24一29
[16]章毓晋.图像工程(下册)一图像理解与计算机视觉[M].北京:清华大学出版社2000年8月.
[17]郑南宁.计算机视觉与模式识别[M],北京:国防工业出版社,2006
[18]杨淑莹,VC++图像处理程序设计[M].北京,清华大学出版社,北京交通大学出版社,2005.1.
[19]求实科技.Visual C++一数字图像处理典型算法及实现[M].人民邮电出版社,2006.6.
[20]李刚,邱尚斌等.基于背景差法和帧间差法的运动目标检测方法[J].仪器仪表学报2006,27(8):64一69.
[21]韩思奇,王蕾.图像分割的闭值法综述[J].系统工程与电子技术.2000,6(24)91一102.
[22]赵荣祷编著.数字图像处理导论.西安:西北工业人学出版社,2000
[23]崔屹.《图像处理与分析一数学形态学方法及应用》.科学出版社,2000
[24] Lipton,and H.Fujiyoshi, etc. Moving target classification and Tracking from real-time video,Proc. IEEE Workshop Applications of Computer Vision,Oct 1998,8(14)
[25] Foresti,G L. Object recognition and tracking for remote video surveillance,IEEE transactions on Circuits and Systems for Video Technology, 1999,9(7):1045一1062
[26] P.Fieguth and D.Terzopoulos,”Color一based Tracking。Heads and other Mobile objects at video frame rates,”In Pore .Of IEEECVPR,pp.21一27,1997.
[27] Ko11er D.Weber J.Towards robust automatic traffic scene analysis in real-time [J」.Proceedings of the IEEE Workshop on Applications of Computer
[28] Zhou S K,Chellappa R, Moghaddan B.Visual tracking and recognition using appearance-adaptive models in particle filters. IEEE Transactions on Image Processing, 2004.
[29]陈朝阳,张桂林.基于图像对称差分运算的运动小目标检测方法[J].华中理工大学学报1998,26(9):34一35
[30]王新余,张桂林.基于光流的运动目标实时检测方法研究[J],计算机工程与应用,2004,l:43一46
[31]冯伟,视频序列运动目标检测与识别方法研究,硕士学位论文,西安,西北工业大学,2003
[32]谭优,王泽勇,图像闭值分割算法实用技术研究与比较[J],微计算机信息,2007,23(24):298一233