Moving targets detection and tracking based on Bayesian foreground segmentation and GVF-snake

We proposed a robust approach to detect and track moving targets observed by a static camera. The approach relies on a Bayes theorem based background model, a GVF-snake based border tracker and a Kalman estimator. The background model is used to segment foreground targets from background, which has the advantages of insensitiveness to initial observations and the capability of adaptive selection of layer number compared with GMM background model. By modifying its energy term and adding automatic initialization of contours, GVF snake is improved to extract the contours of moving targets in video. To speed up convergence, we introduced a Kalman filter to estimate the contour centers. We demonstrated results on a number of different real sequences. The proposed method was proved effective for both rigid and non-rigid objects and can be used for smart surveillance and traffic monitoring.