Performance bounds of extended target tracking in cluttered environments

In this paper, we consider the problem of the performance bound of a nonlinear filtering problem corresponding to tracking an extended target in cluttered environments (i.e., with false alarms and missed detections). The high resolution sensor obtains the measurements of the position and the extent of the extended target whose shape is modeled by an ellipse. The posterior Cramer-Rao lower bound (PCRLB) provides a useful tool to evaluate the best achievable performance of the nonlinear filtering problem. The bounds of the traditional kinematic state estimation are calculated using the point and the extended target model. It is shown in this paper that the bound of extended target tracking is smaller than that of point model because more information is utilized. The bounds are calculated to examine the influence of the measuring accuracy, the geometry between the sensor and the target, the prior knowledge of the target, and the environmental circumstance. In a cluttered environment, the PCRLB is calculated by IRF (information reduction factor), MSC (measurement sequence conditioning), and MESC (measurement existence sequence conditioning) approaches. The simulation results also illustrate the relationship of the three methods.