去相关卡尔曼滤波算法在视频跟踪中的应用  

Application of Decorrelation Kalman Filter in Video Object Tracking

作　　者：戚国庆[1] 陈黎[1] 

机构地区：[1]南京理工大学自动化学院,江苏南京210094

出　　处：《计算机仿真》2008年第11期218-221,共4页Computer Simulation

摘　　要：针对测量噪声去相关下的卡尔曼滤波算法在视频目标跟踪中的应用问题展开了研究,首先分析了图像目标质心测量噪声的统计特性,指出测量误差是具有较强相关性的有色噪声;通过提出一种噪声白化方法,将测量噪声视为在高斯白噪声激励下所产生的一阶马尔科夫序列,并对系统测量模型进行了改进;最后借助噪声去相关意义下的卡尔曼滤波算法对所拍摄的航模飞机视频图像进行了跟踪仿真测试。仿真实验结果证明所给出的滤波算法可以明显提高目标位置的估计精度,并有效抑制了背景的干扰。The decorrerlation Kalman filter used for video target tracking system is studied in this paper. Firstly, the statistic characteristic of the measurement noise in image object centroid position is analyzed, and it is put forward that the measurement error is a color noise which has strong self correlation. Secondly, a color noise whitening method is proposed, which regards the measurement noise as an one - order Markov sequence excited by Gauss white noise. Finally, an experiment of tracking a model airplane in video image is simulated by using decorrerlation Kalman filter. The simulation result proves that the proposed algorithm can improve estimation accuracy obviously, and the background interference is also inhibited.

关 键 词：视频跟踪 有色噪声 卡尔曼滤波 去相关 

分 类 号：TP391.9[自动化与计算机技术—计算机应用技术]