Model parameter adaptive approach of extended object tracking using random matrix

Traditional target tracking technology usually characterizes the target as a point source object. However, this approximation is no longer appropriate when tracking extended objects, such as large size targets and closely spaced group objects. Bayesian extended object tracking (EOT) using random symmetrical positive definite (SPD) matrix is a very effective way to estimate the kinematical state and physical extension of the target jointly. Modeling the physical extension and measurement noise is the key issue when applying this random matrix based EOT approach. In order to improve the performance of extension estimation, model parameter adaptive approaches for both extension evolution and measurement noise are proposed based on the properties of SPD matrix. Some improvements are also made on the prediction formulas and extension dynamic model. Simulation results demonstrate the effectiveness of the proposed adaptive approaches. The estimation error of physical extension is significantly reduced when the target maneuvers.