Video Motion Predictive Tracking Quality: Kalman Filter vs. Lowpass Filter

Predicting the target motion is a common procedure in video target tracking, and Kalman filter has been largely used for this purpose. However, the Kalman filter used as a predictor bears a weakness of compromised prediction accuracy. To tackle this problem, a lowpass filter is engineered in this paper, which is derived from the first-order expansion of Taylor series with incorporation of the inertia. The tests manifest that the proposed lowpass filter possesses much improved predicting capacity than Kalman predictor in terms of prediction precision, and hence may be deemed as a good alternative for motion prediction in video tracking applications.