Abstract:

A new principal component analysis particle filter algorithm  proposed in this paper combines principal component analysis method with the traditional particle filtermethod.  The proposed method avoids undue resampling and improves the accuracy of object tracking.  Experimental results demonstrate that this algorithmcan acquire better precision in single object tracking and multiple\|obstacle object tracking, and this algorithm could track human face accurately in complex backgrounds.  Compared with present particle filter algorithm, this algorithm performs better in robustness and accuracy and avoids particledegeneration and particle impoverishment.

Key words:  object tracking, particle filter algorithm, principal component analysis, resampling