Robust target detection and tracking using particle filter and parametric eigenspace method

In a past, we proposed the image processing method that the tracking of the target object with shape is possible by executing PF and PEM simultaneously. Namely, the PEM is used as a re-sampling method of the PF. And since the PEM is executed at the same frequency as the number of particles, high-speed execution of the proposed method is possible. Moreover, since the posture of the target object can be estimated by using PEM, the position of the moving target object at the next time frame can be estimated in high accuracy. In this paper, it reports that the target tracking is achieved robustly by using estimated posture for prediction part of the PF.