Maneuvering Target Tracking Using Adaptive Models in a Particle Filter

Abstract

Maneuvering target tracking is a big challenge to the performance of a visual tracker. The paper proposes a method to keep the tracker robust to target maneuvering by selecting discriminative features from a large feature space, and constructing a velocity motion model with adaptive noise variance. Furthermore, the feature selection procedure is embedded into the particle filtering process with the aid of calculating the Bhattacharyya distance. Top-ranked discriminative features are selected into the observation model and simultaneously invalid features are removed out to adjust the object representation adaptively. The adaptive motion model is computed via a first-order linear predictor using the previous particle configuration. Experimental results on tracking basketball in video sequences demonstrate the effectiveness and robustness of our algorithm.