Research On Multi-sensor Multi-target Tracking Using Only Range-doppler Measurements

In order to solve the Multisensor-multitarget tracking problem only usingRange-Doppler observations under no angle about targets can be provided by radars,the paper first introduces the principle of generating ghosts in system, and denotesthat the ghosts’ emergence due to the error cross-positioning of the measurementsfrom different targets which were obtained respectively in different radar sensors,then aiming to the characteristics of the ghost tracks the paper analyses theacceleration existing in the ghost tracks’ speed when the true tracks’ speed are notequal with each other, and extractes the characteristic quantity based on the differ ofthe speed squares of two ghost tracks to distinguish the two ghost tracks from truetracks.In the next, the paper presents the filtering and estimation approachs for theradar measurements without the information of targets’ angle. First, the traditionalnonlinear filtering methods such as EKF and UKF are taken respectively to filterand estimate the state of the targets in the system. The simulation results denotesthat the two methods can effectively solve the state estimation problem only usingRange-Doppler measurements in the Multisensor-multitarget tracking system. Inorder to establish the local targets using a single sensor’s measurements, then thepaper presents a Converted Doppler Kalam Filtering approach to construct a pseudostate vector to filter and estimate the local targets’ state.The approach can effectivelysolve the issue that the accelerated motion which is taken in the filtering processusing the traditional state vector consisting of radial distance, Doppler andDoppler-order differential is more rough. And the results of the simulation showsthat the approach is effective.In the end of the paper, two different processes respectively according to thediffer of the speeds from the two targets to remove the ghost tracks in the system arepresented. The system uses CDKF to identify the ghost tracks when the difference ofthe two true targets’ speeds is large, and compares the differ of speeds’ square toidentify the ghost tracks when the difference of the two true targets’ speeds is small.The simulation results denotes that the two approachs are all able to eliminateeffectively the ghost tracks in the system.