基于偏最小二乘分析的双模粒子滤波目标跟踪

Tracking objects based on partial least squares analysis using particle filtering with dual models

针对在复杂背景下,基于主成分分析(PCA)的目标跟踪方法准确率较低的问题,使用偏最小二乘分析,提出一种双模粒子滤波的跟踪算法.首先采用偏最小二乘分析对目标区域建模,作为观测模型;然后利用仿射变换描述目标的形变过程,分别在李群及其切向量空间上建立双模的动态模型;最后结合特征空间更新策略,使用粒子滤波实现目标跟踪.实验表明,所提出的算法能够有效滤除背景噪声,跟踪结果稳定且准确.

For the problem that the object tracking algorithm using principal components analysis（PCA） has low accuracy in a complex environment, based on the partial least squares analysis, an object tracking algorithm is proposed by using particle filtering with dual models. Firstly, the model of object region is built by the partial least squares analysis, which is applied as the observation model. Then, the dynamic model with dual models is built on Lie group and the corresponding tangent vector space respectively, with the describing the object deformation process by affine transformation. Finally, combining with the update strategy for feature space, the object tracking algorithm is realized by particle filtering. Experiments show that the tracking results are stable and accurate, and the proposed algorithm can effectively filter out the background noise.