低信噪比扩展目标跟踪方法研究

发布时间：2018-03-13 20:09

本文选题：目标跟踪　切入点：IMM-TBD　出处：《西安电子科技大学》2014年硕士论文　论文类型：学位论文

【摘要】：雷达数据处理的主要功能是完成对目标的跟踪。经典的目标跟踪算法以点目标假设为前提,通过航迹起始、点航关联及状态滤波等操作来实现航迹处理,但是随着雷达分辨率的不断提高,目标量测将会分布在多个分辨单元内,采用传统的跟踪算法对扩展目标进行跟踪时会面临数据关联复杂度高及跟踪发散等难题,有必要研究适用于扩展目标的跟踪方法。本文首先介绍了点目标跟踪基础理论,包括经典的航迹起始算法、数据互联算法及滤波算法。其中航迹起始算法主要有逻辑法和Hough变换法;数据关联算法主要有最近邻域互联算法、强近邻域互联算法、概率数据互联算法以及联合概率数据互联算法;滤波算法主要有卡尔曼滤波、扩展卡尔曼滤波、不敏卡尔曼滤波以及交互式多模型滤波算法。结合仿真实验,对比分析了上述典型算法的性能。然后对比研究了噪声背景下的两种点目标跟踪算法,其中一种是线性非高斯系统中的序贯贝叶斯估计方法,该方法具有高斯和滤波算法的优点,并通过引入模型阶数降低步骤解决了高斯和滤波算法模型阶数呈指数型增长的问题。接着重点提出了一种基于轨迹增强的IMM-TBD算法,采用一组增强算子对目标轨迹进行增强检测,并将该算子与交互多模型算法有效结合,从而解决了低信噪比情况下高机动目标的跟踪问题。最后对比研究了两种基于概率假设密度(PHD)的扩展目标跟踪算法,其中一种算法基于随机有限集理论,该算法将扩展目标的量测集合建模为随机有限集,适用于杂波背景下非邻近多目标的跟踪。另一种算法为基于随机矩阵的PHD滤波方法,该算法将目标的扩展情况建模为随机矩阵,适用于杂波背景下邻近多目标的跟踪。
[Abstract]:The main function of radar data processing is performed on the target tracking. Target tracking algorithm with the classic point target assumption for the track initiation point, association and filter operations to achieve track processing, but with the improvement of the resolution of radar target, the measurement will be distributed in more than one resolution cell, by tracking the traditional algorithm of extended target tracking data association will face high complexity and tracking divergence problem, it is necessary to study the method of tracking for extended targets. This paper first introduces the basic theory of point target tracking, including track initiation algorithm, data association algorithm and filtering algorithm. The algorithm of track initiation are logical method and Hough transform method; nearest neighbor association algorithm main data association algorithm, strong neighborhood association algorithm, probabilistic data association algorithm and joint probability Data association algorithm; Calman filter is the main filter algorithm, extended Calman filter, unscented Calman filter and interacting multiple model algorithm. With the simulation results, comparative analysis of the performance of the typical algorithms. And comparative study of the two kinds of point target noise background tracking algorithm, one of which is a sequential Bayesian linear non Gauss system the estimation method, this method has the advantages of Gauss and filtering algorithm, and by introducing the model order reduction steps to solve Gauss and filtering algorithm for model order exponential growth. Then we propose a IMM-TBD algorithm based on enhanced trajectory, using a set of enhanced operator to track the target to enhance detection and, the effective combination of the operator and the interactive multi model algorithm to solve the tracking problem of low SNR and high maneuvering target. Finally, comparative study Two hypotheses based on probability density (PHD) of the extended target tracking algorithm, an algorithm based on the theory of random finite set, the algorithm will be extended target measurement set is modeled as random finite sets, suitable for clutter non adjacent targets tracking. Another algorithm for PHD filtering method of random matrix based on the algorithm will be extended to the modeling target for random matrix, suitable for clutter adjacent targets tracking.

【学位授予单位】：西安电子科技大学
【学位级别】：硕士
【学位授予年份】：2014
【分类号】：TN953

【参考文献】