摘要:本文研究的是多目标视频跟踪技术研究。首先,简单全面地介绍了多目标视频跟踪系统,简要论述了跟踪的起始与终结,跟踪文持过程,跟踪门的设计和数据关联的概念与作用。然后,对数据关联算法进行了深入的研究。对最近邻数据关联(NNDA)、概率数据关联(PDA)、联合概率数据关联(DA)和交互式多模型(IMM)这四种经典算法进行深入剖析和理论仿真,并分析、比较了其适用范围和跟踪性能。最后,为了提高跟踪性能,在DAF的基础上引入目标区域颜色直方图这个匹配因素,提出了融合颜色直方图的DAF算法,并将该算法应用到视频图像中的多目标跟踪中。实验结果表明,该算法能对视频图像中的多个运动目标进行有效的跟踪。35505
毕业论文关键词:视频图像 多目标跟踪 数据关联 颜色直方图
Multi-target tracking in video
ABSTRACT:This dissertation studies the data association algorithm based on
multi-target tracking in video image.Firstly, a brief introduction to the entire system of multi-target tracking in video image. The tracking formation and termination, the process of tracking maintenance, the design of tracking gate and the concept and role of data association are briefly introduced.
Secondly, the data association algorithms is in-depth researched. the four classical algorithms which include the nearest neighbor data association (NNDA), probabilistic data association (PDA), joint probabilistic data association (DA) and interactive multi-model (IMM) are deep studied and theoretically simulated, comparing and analysing their applicable range and tracking performance.
Finally, in order to improve the tracking performance, we consider the color histogram of target area on the basis of DA and we propose a new kind of algorithm which is DA algorithm based on the color histogram. In the end , we apply the new algorithm to multi-target tracking in video image. Experimental results show that this algorithm can effectively track multiple moving targets in video image.
Key words: Video Image Multi-target Tracking Data Association Color Histogram
目录
1 绪论 1
1.1 研究背景与意义 1
1.2 本文主要内容 1
2 多目标跟踪及其数据关联的基本理论 2
2.1 多目标跟踪的基本理论 2
2.1.1 目标跟踪系统介绍 2
2.1.2 跟踪起始和终结 2
2.1.3 跟踪门技术 3
2.1.4 滤波和预测 3
2.2 多目标跟踪中的数据关联 3
2.3 目标运动模型 4
2.3.1 CV模型 4
2.3.2 CA模型 4
2.3.3 转弯模型 5
2.4 本章小结 5
3 数据关联算法及其仿真 6
3.1 卡尔曼(kalman)滤波 6
3.2 几种经典的数据关联算法 7
3.2.1 最近邻域法 7
3.2.2 概率数据关联算法 7
3.2.3 联合概率数据关联算法 9
3.2.4 交互多模型算法 11
3.3 仿真结果和分析 14
3.3.1 最近域法与概率数据关联算法 14
3.3.2 联合概率数据关联算法 16 JPDAF算法视频图像多目标跟踪技术研究:http://www.youerw.com/tongxin/lunwen_33493.html