Prediction-assisted moving objects tracking in severe occlusions and background clutters

Abstract

In this paper, a novel approach to multi-object tracking is proposed. In this approach a modified MSA along with a new adaptive Kalman filter design and foreground correlation tracker is proposed. The main contributions of this work are three. First, a new adaptive process noise and measurement noise is designed for the Kalman filter. Second, a foreground correlator based on background subtraction is designed to track the target undergoing abnormal motion and prolonged occlusion. Third, a normalized RGB color space is considered in the proposed method in order to make the tracking system less sensitive to lighting changes and also can considerably reduced by the amount of data to process.