Abstract
An adaptive particle filter with a statistical motion model, called the optical flow--driven motion model, is proposed for sequential Bayesian visual tracking. The motion model predicts the current state with the help of optical flows, i.e., it explores the state space with information based on the current and previous images in an image sequence. This exploration improves the prediction accuracy, compared to a prior model based exploration, which is widely used in visual tracking. In addition, an automatic method for adjusting the variance of the motion model is introduced. The variance affects the performance of tracking but the parameter is manually determined in most particle filters. In experiments with two real image sequences, the proposed motion model is compared with a random walk model. The experimental results show the proposed model outperform the random walk model in terms of accuracy even though their execution times are almost the same.
