一种改进的分布式多目标跟踪算法

将多平台分布式的IMM_Kalman(DIMM_Kalman)算法和IMM_JPDA(DIMM_JPDA)算法相结合,得到改进的DIMM_JPDA(DIIMM_JPDA)算法。在DIIMM_JPDA算法中,每个平台都有自己独立的IMM算法跟踪系统,每个IMM算法中都采用Kalman/JPDA滤波器,全局的滤波值由各平台的滤波值融合而成。然而在全局融合中,由于协方差矩阵可能出现奇异,导致计算的不稳定性或出现严重异常。为了克服上述问题的缺点,引入了特征值扰动方法,使该协方差矩阵变成非奇异矩阵。仿真实验表明改进的DIIMM_JPDA算法比DIMM_JPDA算法有更小的误差。

Improved Distributed Multi-target Tracking Algorithm

Distributed interacting multiple Kalman_model(DIMM_Kalman) and JPDA_model(DIMM_JPDA) fusion algorithm for multi-platform tracking were developed to be a distributed interacting multiple JPDA-model(DIIMM_JPDA).Locally,each platform contains its own independent tracking subsystem,using an IMM estimation algorithm,and an IMM algorithm contains some Kalman/JPDA filters.Globally,equivalent platform models and an equivalent global model are constructed so that a global multiple model fusion algorithm can be carried out.In the global fusion,the covariance matrix may be singular,so the calculation of the instability or severe abnormalities.To overcome the shortcomings of the above problems,the eigenvalue perturbation method was introduced,to makethe covariance matrix into a non-singular matrix.In this case,compared with a DIMM_JPDA algorithm,the DIIMM_JPDA algorithm has less mean-errors.