A Study Of Random Finite Set Based Intensity Filter Under Unknown Clutter Environment

In recent years, multi-target tracking technology is increasingly highlighting its important role in the national defense and civil applications. Due to the detection uncertainty and association uncertainty, its development in theoretical research and practical applications is still facing a relatively big challenge. The clutter intensity, which is often difficult to be known as a priori, is a key parameter affecting the performance of target tracking. Therefore how to achieve the effective and accurate estimation of the clutter intensity of the tracking environment is a subject worthy of research. This dissertation focuses on the applications of the intensity filter(Intensity filter, i Filter) that implement based on the random finite set(Random finite set, RFS)theory under the unknown clutter environment. The main contributions of the dissertation are as follows:1, The RFS filter model and principles as well as two basic implement methods of the probability hypothesis density(Probability Hypothesis Density, PHD)filter based on this model: Gaussian mixture probability hypothesis density(GM-PHD)filter and Sequential Monte Carlo probability hypothesis density(SMC-PHD)filter are introduced in this dissertation.2, Aiming at solving the unknown clutter intensity problem, researches on the particle intensity filter that implements based on the RFS theory have been done. The dissertation first introduces the intensity filter model which is implemented based on the Poisson point processes(Poisson Point Processes,PPPs)and its implement processes, followed by introducing the augmented state space to the RFS multi-target tracking model to construct the model of the conversion between the targets and the clutter, thus the SMC-i Filter algorithm that implement based on the RFS theory is introduced. This algorithm can achieve the clutter intensity at each time accurately and in real-time, which effectively solves the tracking performance reducing issues caused by the unknown clutter intensity. It concludes with proposing the marginalized particle intensity filter(MP-i Filter) which aims at overcoming the problem of underestimate on targets number in SMC-i Filter. In this algorithm, the marginalized particle filter algorithm is introduced to approximate the target particle set, enhancing the samplingefficiency by reducing the dimension of the sampling space and hence further improve the filtering performance.3, In order to improve the performance of the estimate on target position during the tracking process, Forward-Backward smoothing algorithm is introduced firstly. Then a particle intensity smoothing algorithm is proposed by combining the Forward-Backward smoothing algorithm with the intensity filter. The algorithm proposed makes a good correction on the estimate of the target position and thus improves the space position estimate performance. The Forward-Backward smoothing algorithm is also used in the marginalized particle intensity filter and therefore improve the filtering performance on both the target number estimate and the position estimate. The simulation results demonstrate the effectiveness of the algorithm.