Monte Carlo data association for multiple target tracking

The data association problem occurs for multiple target tracking applications. Since non-linear and non-Gaussian estimation problems are solved approximately in an optimal way using recursive Monte Carlo methods or particle filters, the association step will be crucial for the overall performance. We introduce a Bayesian data association method based on the particle filter idea and the joint probabilistic data association (JPDA) hypothesis calculations. A comparison with classical EKF based data association methods such as the nearest neighbor (NN) method and the JPDA method is made. The NN association method is also applied to the particle filter method. Multiple target tracking using particle filter will increase the computational burden, therefore a control structure for the number of samples needed is proposed. A radar target tracking application is used in a simulation study for evaluation.