Measurement Combination Estimator for Multisensor Extended Object Tracking Using Random Matrix

For extended object tracking (EOT), the random-matrix approach has received extensive attention. It is able to jointly estimate the kinematic state and extension of an extended object. Multiple sensors with different characteristics and perspectives can provide new information especially for EOT. However, how to fuse the measurements from multiple sensors to guarantee the improvement of the estimation accuracy of the state and extension still needs further research. This article proposes a random-matrix approach to elliptical EOT using multiple sensors based on a centralized fusion architecture. In order to handle the different distortions of multiple sensors, a measurement combination (MC) approach is first proposed. As a result, a likelihood function with a proper form is obtained, and it can also preprocess the multisensor measurements and take the scattering matrix of measurement means into consideration. Based on the likelihood function, the MC estimator (MCE) is derived for multisensor EOT in a Bayesian framework. It inherits the analytical and concise scheme of the single-sensor EOT approach. Based on the proposed MCE approach, it is theoretically proved that the estimation accuracy of the state and extension can be improved when the sensor number increases. Effectiveness of the MCE approach is demonstrated by simulation and experimental results compared with the existing multisensor EOT approaches using random matrix.