摘要
针对室外条件下动态纹理背景,采用自回归运动平均(ARMA)模型建立背景模型,并引入快速增量主元分析(IPCA)算法对模型进行降维,并辨识其中参数,实现最大似然估计。运用增量主元分析算法,不需要估算协方差矩阵,直接可以递增地得到特征向量和奇异值,计算出样本序列的主要元素。完成参数辨识后,ARMA模型可以合成无限长度的预测图像序列。最后,仿真实验证明了算法的有效性。
A dynamic textured background in real world situations was modeled by an Antoregressive Moving Average (ARMA) model. Then a fast Incremental Principal Component Analysis (IPCA) algorithm was introduced to reduce dimensionality and identify, and compute the principal components of a sequence of samples incrementally without estimating the covariance matrix. Once learned, a model had predictive power and can be used for extrapolating synthetic sequences with infinite length. Preliminary experiments with this method have achieved promising results.
出处
《计算机应用》
CSCD
北大核心
2009年第B12期241-243,共3页
journal of Computer Applications
关键词
动态纹理
背景建模
自回归运动平均模型
增量主元分析
子空间系统辨识
dynamic textures
background modeling
Autoregressive Moving Average (ARMA) model
Incremental Principal Component Analysis (IPCA)
subspace system identification