Measurement partition algorithm based on density analysis and spectral clustering for multiple extended target tracking

Abstract

For the multiple extended target tracking (METT), one crucial problem is how to partition the measurement sets accurately and rapidly. Due to the disturbance of clutter, the conventional methods, such as distance partition method, K-means++ method, etc., cannot give a perfect partition. In this paper, a novel partition method is proposed based on density analysis and spectral clustering technique. Firstly, construct the density distribution function of the measurements by using the Gaussian kernel, and then eliminate the clutter from the measurements. Secondly, the spectral clustering technique based on neighbor propagation is introduced to partition the measurements. Finally, the Gaussian mixture probability hypothesis density method is used to achieve the METT. Simulation results show that the proposed algorithm has a better performance, especially a better real-time performance, than the conventional methods.