An Efficient Object Tracking Algorithm with Adaptive Prediction of Initial Searching Point

In object tracking, complex background frequently forms local maxima that tend to distract tracking algorithms from the real target. In order to reduce such risks, we utilize an adaptive Kalman filter to predict the initial searching point in the space of coordinate transform parameters so that both tracking reliability and computational simplicity is significantly improved. Our method tracks the changing rate of the transform parameters and makes prediction on future values of the transform parameters to determine the initial searching point. More importantly, noises in the Kalman filter are effectively estimated in our approach without any artificial assumption, which makes our method able to adapt to various target motions and searching step sizes without any manual intervention. Simulation results demonstrate the effectiveness of our algorithm.