Performance comparison of particle swarm optimization and extended Kalman filter methods for tracking in non-linear dynamic systems

The tracking is an application of computer vision has grown remarkably over the last two decades. By tracking, exist a set of algorithms that provide detecting and tracking objects in a video sequence. It then touches several areas, such as video surveillance biometric or biomedical imaging, the human-machine interaction, traffic control and intelligent vehicles, etc. The multiplicity of problems to solve makes the tracking a challenge to researchers who offer nonlinear / non-Gaussian systems to increase either the quality of tracking either the speed of the convergence time. In this context, several authors proposed different approaches for the tracking in a dynamic application. Extended Kalman filter is one of the popular techniques for the tracking and is useful when compared to some of the other tracking approaches. Otherwise, this method has a drawback present by the linearizing of nonlinear system dynamics and observation models may induce errors in the estimation of the state. Particle swarm optimization method used for solving a complex problem and a tracking many of applications but this algorithm has low convergence time. This paper presents a comparison of both of these algorithms for object tracking in nonlinear dynamic systems. Observations based on the methods mentioned are present in the literature but there is no evidence based on the applications of these algorithms that show a quantitative comparison for the best algorithm.