Study of Multi-Target Tracking Algorithm Based on Random Sets
|School||Xi'an University of Electronic Science and Technology|
|Course||Control Theory and Control Engineering|
|Keywords||Multi-target tracking Random sets Probability Hypothesis Density Measurement noise|
As a typical signal processing problems , target tracking has been widespread concern , and has broad application prospects . With the application of the complex, simple, single- target tracking has been difficult to meet the demand, so multi-target tracking has become the focus of research scholars at home and abroad . Traditional multi- target tracking data association algorithm as the core, its tracking performance is greatly affected by the data association . Unlike this , based on random set of multi-target tracking algorithm does not require data association, effectively overcome the shortage of traditional methods , is to track hot research field , but also the focus of this study . First , this paper introduces the basic theory and the tracking filter several classic filtering algorithm, and the simulation results are analyzed and compared its performance . Then, expounded the basic principle of multi-target tracking , and on this basis, introduced the tradition based on data associated with multi-target tracking algorithm . In this paper, based on random set of multi-target tracking method with a focus on its principles and classical algorithms conducted in-depth research and analysis carried out by simulation verification . Finally, classical probability hypothesis density filter algorithm shortcomings , this paper presents several improved algorithms. First, the particle probability hypothesis density filter for particle updates during the measurement noise distribution dependent , this paper proposes an unknown measurement noise distribution under the multi- target tracking algorithm . Second, for a Gaussian particle probability hypothesis density filter in the prediction and update repeated particle approximation and sampling , this paper proposes an improved Gaussian particle probability hypothesis density filter. Third, assuming Gaussian probability density for mixed tracker can track multiple moving targets , this paper proposes a method to track multiple moving targets with a mixture of Gaussian functional probability hypothesis density tracker algorithms. In this paper, simulation results show that : the original algorithm, the improved algorithm has better tracking performance.