Occlusion detection and object tracking using filter banks

When objects are occluded they are obtruded from human vision. A new means has been proposed to track such occluded objects. This paper proposes a novel approach to detect and track occluded moving objects. A Kalman Filter Bank (KFB) approach has been used to predict the motion of the moving targets using the current state and the previous few states. Based on the linearity level the corresponding filters have been used. When the moving object is linear then linear Kalman filter is used, if the motion is non-linear then Extended Kalman Filter (EKF) is used and Unscented Kalman Filter (UKF) is used in case of high non-linearity. Kalman Filter uses measurements that are observed over time and produce values that tend to be closer to the actual values. The EKF linearizes about the current mean and covariance using a Taylor series based transformation. The UKF utilizes a set of sample points, which guarantees accuracy with the posterior mean and covariance to the second order for any nonlinearity. It predicts the next state by deriving estimates from the previous states. The proposed work varies from others by incorporating a cycle from correction state to the prediction state which provides a greater advantage in refining the expected state value. For an experimental evaluation, sequences comprising of continuous data are used.