积分粒子滤波方法及其应用研究

发布时间：2018-09-19 17:56

【摘要】：目前,大规模被动传感器系统及其相关关键技术的研究日益受到国内外学者的重视。针对这种观测信息有限、数据丢失率高,具有非线性、非高斯特征的被动传感器观测数据的滤波处理是大规模被动传感器系统数据处理首先要解决的关键和难点问题。结合自已有非线性非高斯滤波算法,本文提出了Gauss-Hermite积分的高斯和积分粒子滤波算法以及考虑目标特性的辅助高斯和积分粒子滤波算法,在此基础上,给出了适合并行计算的并行高斯和积分粒子滤波算法。针对大规模被动传感器系统中非线性非高斯观测数据的滤波问题,提出了基于Gauss-Hermite积分的高斯和积分粒子滤波器(GSQPF)。GSQPF利用积分点概率密度度函数作为重要性密度函数,同时利用高斯混合更新后验概率,有效的提高了采样粒子的多样性与准确性。仿真结果表明,提出的高斯和积分粒子滤波估计性能明显要好于高斯和粒子滤波(GSPF)、积分粒子滤波(QPF),可以对非线性非高斯系统进行精确的状态估计。为增强对非周期稀疏采样观测数据的滤波处理能力,在GSQPF的基础上,将时间间隔、目标观测和目标速度等目标特性融入到重要性密度函数的构建中,提出了一种辅助的高斯和积分粒子滤波(AGSQPF),有效的增强了采样粒子的多样性和准确性。实验结果表明,提出的AGSQPF估计精度要优于GSQPF方法,能够对被动目标进行准确跟踪。针对大规模被动传感器系统观测数据处理量大、通讯需求高的问题,给出一种并行高斯和积分粒子滤波算法。在算法的并行处理上,高斯和积分粒子滤波器的粒子和权值都是在子系统进行更新,各子系统都考虑了边界状态信息,但各子系统的滤波过程是相互独立的,大大提高了传感器系统数据处理的效率。由于实验结果表明,提出的算法能够满足大规模被动传感器系统应用的需要,能够对目标进行准确跟踪。
[Abstract]:At present, the research of large-scale passive sensor system and its related key technologies has been paid more and more attention by scholars at home and abroad. In view of the limited observation information, high data loss rate, nonlinear, non-Gao Si characteristics of passive sensor observation data filtering processing is the first to solve the key and difficult problem of large-scale passive sensor system data processing. Combined with their own nonlinear non-Gao Si filtering algorithm, this paper proposes the algorithm of Gauss-Hermite integral, which is Gao Si and integral particle filter, as well as the auxiliary Gao Si and integral particle filter algorithm, which takes into account the characteristics of the target. On the basis of this, The parallel Gao Si and integrated particle filter algorithms suitable for parallel computation are presented. Aiming at the filtering problem of nonlinear non-Gao Si observation data in large-scale passive sensor systems, a new method based on Gauss-Hermite integral is proposed, which uses the integral point probability density function as the importance density function, and the integrated particle filter (GSQPF). GSQPF uses the integral point probability density function as the importance density function. At the same time, the diversity and accuracy of sampling particles are improved effectively by using Gao Si mixed update posterior probability. The simulation results show that the performance of Gao Si and integrated particle filter is better than that of Gao Si and (GSPF), integrated particle filter (QPF),. In order to enhance the filtering and processing ability of aperiodic sparse sampling data, based on GSQPF, target characteristics such as time interval, target observation and target velocity are incorporated into the construction of importance density function. An auxiliary Gao Si and integrated particle filter (AGSQPF),) are proposed to effectively enhance the diversity and accuracy of sampling particles. The experimental results show that the proposed AGSQPF estimation method is more accurate than the GSQPF method and can track the passive target accurately. A parallel Gao Si and integrated particle filter algorithm is proposed to solve the problem of large data processing and high communication demand in large-scale passive sensor systems. In parallel processing of the algorithm, the particles and weights of Gao Si and integral particle filter are updated in the subsystem, each subsystem considers the boundary state information, but the filtering process of each subsystem is independent of each other. The efficiency of data processing in sensor system is greatly improved. The experimental results show that the proposed algorithm can meet the needs of large-scale passive sensor system applications and can accurately track the target.
【学位授予单位】：深圳大学
【学位级别】：硕士
【学位授予年份】：2015
【分类号】：TN713;TP212.9

【参考文献】