Tracking objects in video sequence using active contour models and unscented Kalman filter

In this paper, we propose a new association of active contour model (ACM) with the unscented Kalman filter (UKF) to track deformable objects in a video sequence. The proposed approach is based on the use of the selective binary and Gaussian filtering regularization level set associated to the UKF (ACM-SBGFRLS-UKF) instead of the traditional level set (TLS) associated to the UKF (ACM-TLS-UKF). In fact, in the present work, we exploit the various advantages that the SBGFRLS offers compared to the TLS which suffers, from the sensibility to initials conditions and noise, to the impossibility to select partial or global segmentation and also from the complexity of the approach. Finally, a comparison study is presented, throughout several numerical simulations, of this new association approach ACM-SBGFRLS-UKF against the ACM-TLS-UKF for tracking deformable objects in a video sequence.