Multiple Target Tracking Using Cheap Joint Probabilistic Data Association Multiple Model Particle Filter in Sensors Array
Zahir Messaoudi · International Journal of Artificial Intelligence & Applications · 2012
Joint multiple target tracking and classification is an important issue in many engineering applications.In recent years, multiple sensor data fusion has been extensively investigated by researchers in a variety of disciplines.Indeed, combining results issued from multiple sensors can provide more accurate information than using a single sensor.In the present paper we address the problem of data fusion for joint multiple maneuvering target tracking and classification in cluttered environment where centralized versus decentralized architectures are often opposed.The proposal advocates a hybrid approach combining a Particle Filter (PF) like method to deal with system nonlinearities and Fitgerald's Cheap Joint Probabilistic Data Association Filter CJPDAF for the purpose of data association and target estimation problems, yielding CJPDA-PF algorithm.While the target maneuverability is tackled using a combination of a Multiple Model Filter (MMF) and CJPDAF, which yields CJPDA-MMPF algorithm.Consequently, at each particle level (of the particle filter), the state of the particle is evaluated using the suggested CJPDA-MMF.In case of several sensors, the centralized fusion architecture and the distributed architecture in the sense of Federated Kalman Filtring are investigated and compared.The feasibility and the performances of the proposal have been demonstrated using a set of Monte Carlo simulations dealing with two maneuvering targets with two distinct operation modes and various clutter densities.