A particle swarm optimization approach for multi-objects tracking in crowded scene

This paper presents a new particle swarm optimization-based algorithm for tracking objects in crowded scenes. The proposed method exploits the properties of local feature descriptors and color-based covariance matrix to model the targets. Then, optimal search for the best match of the targets in the successive frames is performed using a particle swarm optimization (PSO) algorithm. The PSO, which is a population-based searching algorithm, attracts all particles towards the global optima based on a fitness function defined using a color-based covariance matrix. Adaptation of tracking windows is obtained based on local feature descriptors. Local feature descriptors are extracted using the scale invariant feature transform (SIFT) method. Our proposed method can cope with a number of challenging scenarios typical of crowded scenes. This includes tracking objects under heavy occlusions, erratic motion and illumination changes.