Research On Random Finite Sets Tracking With Adaptive Target Birth Intensity

Multi-target tracking is an important and d ifficult topic in information fusion. Because of its high application value both in military and civil areas, multi-target tracking technology always attracts much attention by researchers and engineers worldwild. With the study progress of multi- target tracking methods based on random finite sets theory, multi-target tracking research has made a significant advance rapidly.Early multi- target tracking methods assume that the intensity of newborn targets' random set is known a priori. However, in a real complex scenario, targets' birth intensity is quite difficult to obtain in advance. Therefore, with the unknown newborn targets' intensity, we need a reliable method to track multiple targets stably. This thesis mainly focuses on the various multi-target track methods with the unknown targets' birth intensity in the framework based on random finite sets theory. The main work is as follows:Firstly, an overview of the basic concepts of random finite sets theory and some co rrelate filtering algorithms ha ve been provided. The probability hypothesis density filter(PHD) and cardinalized probability hypothesis density filter(CPHD) based on random-set techniques have been introduced in details, and their recursive steps of Gaussian Mixture solutions have been given.Secondly, the traditional targets' Gaussian mixtures(GM) newborn model has been introduced. To overcome the drawback of GM model, an adaptive target birth intensity PHD filter has been researched in details. For the contradiction in the detection step of newborn targets and clutter, an estimate method of target birth rate, which can reduce the clutter's impact on newborn target detection, has been introduced. The confirm of newborn targets' appearance time is always delayed when clutter exists, which may make an negative impact on subsequent process, such as track correlation, therefore an adaptive target birth intensity PHD smoother has been proposed. This algorithm combines backward smoothing method with the estimate method of target birth rate. The analysis and simulation results show that the algorithm is able to estimate the state and get the moment of newborn targets' appearance more accurately and achieve a better track performance.Finally, adaptive target birth intensity CPHD filtering algorithm has been researched. The analysis and comparison between adaptive target birth intensity CPHD filter and adaptive target birth intensity PHD filter have been made. Under the condition of unknown clutter rate, an improved algorithm for adaptive target b irth intensity CPHD filter with unknown clutter rate has been proposed, whose implemented steps by Gaussian Mixture form have been given. This filter can track multiple targets stably when clutter rate and target birth intensity are both unknown. It not only needn't the target birth intensity is known a priori, but also can estimate the clutter rate of the scene on- line. Simulation results show that the proposed algorithm is effective and practical.