Adaptive probabilistic tracking with reliable particle selection

A novel, effective probabilistic tracking method is proposed to adaptively capture the varying target appearance in a complex environment. Different from the traditional particle filter algorithms, the proposed method estimates the weight of each particle not only through similarity measurement between the target model and each hypothetical observation, but also through dissimilarity measurement between the background model and each hypothetical observation. The reliable particles with high weights are then selected to estimate the target state, and the target model is evolved over time with a novel model update strategy. Comparison experimental results demonstrate the robust performance of the proposed algorithm under challenging conditions.

[1]  Takahiro Ishikawa,et al.  The template update problem , 2004, IEEE Transactions on Pattern Analysis and Machine Intelligence.

[2]  Rama Chellappa,et al.  Visual tracking and recognition using appearance-adaptive models in particle filters , 2004, IEEE Transactions on Image Processing.

[3]  Thomas S. Huang,et al.  Online updating appearance generative mixture model for meanshift tracking , 2007, Machine Vision and Applications.

[4]  Dorin Comaniciu,et al.  Kernel-Based Object Tracking , 2003, IEEE Trans. Pattern Anal. Mach. Intell..

[5]  Luc Van Gool,et al.  An adaptive color-based particle filter , 2003, Image Vis. Comput..

[6]  Bohyung Han,et al.  Visual Tracking by Continuous Density Propagation in Sequential Bayesian Filtering Framework , 2009, IEEE Transactions on Pattern Analysis and Machine Intelligence.

[7]  Michael Isard,et al.  CONDENSATION—Conditional Density Propagation for Visual Tracking , 1998, International Journal of Computer Vision.