雷达机动目标运动模型与跟踪算法研究

发布时间：2019-05-22 19:12

【摘要】：目标跟踪问题是一个随着跟踪对象的变化、发展而不断发展、深入研究的问题。通过目标跟踪，实现对目标状态的精确估计，从而为后续的很多信息处理，如目标威胁估计、指挥决策等提供稳定的数据基础。由于新型跟踪目标的出现以及对目标跟踪信息的不断需求，机动目标跟踪越来越成为当前研究热点。论文结合863课题:“空天多源信息×××研究”，主要开展雷达机动目标的运动建模与滤波跟踪算法方面的研究。论文的主要内容包括：首先介绍了论文的研究背景，并对机动目标跟踪中的两大问题：目标运动模型、跟踪算法的研究现状进行了详细论述，并介绍了本文的研究内容。以参数“α”和“η”为特征参量，建立了基于α-η参数的强机动目标运动模型。通过详细分析Singer模型和Jerk模型的特征，分析了二者在表征目标运动特征方面的不足。基于此，，以参数α和η为参数特征，建立了强机动目标的α-η参数运动模型。通过对α-η参数运动模型的离散化处理，推导出α-η参数运动模型的状态-测量模型，并详细分析了α-η参数运动模型的特征。实验表明该运动模型具有较强的目标机动模式表征能力。提出了一种基于修正不敏卡尔曼滤波的目标跟踪算法。在UKF算法中，滤波增益的计算主要由两个协方差决定：状态协方差、状态与测量的协方差，当目标作机动时，滤波增益将滞后于目标的机动状态，从而使跟踪误差变大。因而，在跟踪过程中，通过实时估计噪声协方差的修正因子，然后利用修正因子实时修正预测状态协方差，利用修正后的预测协方差更新状态协方差，进而修正滤波增益。采用自适应因子修正后的协方差来计算滤波增益，使得修正后的滤波增益与目标的运动相匹配，从而获得较好的滤波跟踪精度。实验表明该算法具有比UKF更好的跟踪性能。融合UT变换和EKF各自优点，在提高算法的跟踪性能和较少运算时间方面，提出了两种目标跟踪算法。（1）不敏扩展卡尔曼滤波跟踪算法。UKF在非线性跟踪系统中具有比EKF更好的跟踪性能，但是所需的计算时间大于EKF的计算时间。基于此原因，提出了一种融合不敏变换（UT）和扩展卡尔曼滤波的目标跟踪方法，该方法主要通过把UKF中状态协方差以及状态和测量值的互协方差的多项矢量相乘转换成多个相加的计算，从而有效减少算法的运算时间。该算法融合UT变换的多样性Sigma粒子的特点以及EKF的运算时间快的特点，既保留了较好的滤波跟踪精度，又具有较少的运算时间。（2）自适应不敏扩展卡尔曼滤波跟踪算法。在不敏扩展卡尔曼滤波过程中，利用残差信息，采用指数衰减和遗忘因子的方式实时估计和修正两个噪声协方差，实现噪声协方差的自适应估计。实验表明这两种算法具有比UKF较好的跟踪精度，又具有较少的运算时间。在提高模型概率估计准确性方面，提出了一种基于模型概率修正的交互多模型算法。交互多模算法在计算滤波后的状态信息时，加权因子（即模型概率）的计算主要利用两类信息：新息和模型概率预测值，该方法没有利用当前时刻状态协方差的有效信息，造成对模型概率估计的不准确。基于这个特性，把状态协方差的信息融合得到另一个加权因子，利用该加权因子修正IMM算法中的模型概率估计值，即：加权因子的模型概率修正。该算法既利用了预测模型概率因子，又利用了当前状态方差加权因子，因而，具有较为准确的模型选择概率估计。通过实验验证了该算法具有比IMM较准确的模型概率估计能力。最后对论文的工作进行了总结，并指出论文的不足和今后的研究方向。
[Abstract]:The problem of target tracking is a problem that is developed and researched deeply with the change and development of the tracking target. Through the target tracking, the accurate estimation of the target state is realized, so that a stable data base is provided for a plurality of subsequent information processing, such as the target threat estimation, the command decision and the like. As a result of the emergence of the new tracking target and the continuous demand for the target tracking information, the tracking of the maneuvering target is becoming the focus of the current research. In this paper, the research on the motion modeling and the filter tracking algorithm of the radar maneuvering target is mainly carried out in the "A Study on the Multi-source Information of the Sky-sky" of 863. The main contents of the thesis include: In this paper, the research background of the paper is introduced, and the two main problems in the tracking of the maneuvering target are introduced: the target motion model, the research status of the tracking algorithm are discussed in detail, and the research in this paper is also introduced. On the basis of the parameter "a hand" and the "a hand" as the characteristic parameters, the strong maneuvering target transportation based on the parameter of the parameter of the parameter is established. The characteristics of the Singer model and the Jerk model are analyzed in detail, and the characteristics of the target motion are analyzed by analyzing the characteristics of the Singer model and the Jerk model. In this paper, the parameters of the strong maneuvering target are set up based on the parameters and the parameters, and the parameters of the strong-maneuvering target are set up. In this paper, the state-to-measure model of the motion model of the P-P parameter is derived by the discretization of the motion model of the P-P parameter, and the motion model of the motion model is analyzed in detail. The experimental results show that the motion model has a strong target maneuver model. The purpose of this paper is to provide an object based on the modified non-sensitive Kalman filter. in a UKF algorithm, the calculation of the filter gain is mainly determined by two covariance decisions: state covariance, state and measured covariance, and when the target is mobile, the filter gain will lag behind the maneuvering state of the target, so as to The tracking error is increased. Thus, in the tracking process, the correction factor of the noise covariance is estimated by real-time, then the prediction state covariance is corrected in real time by the correction factor, the state covariance is updated with the modified prediction covariance, the filter gain is calculated by using the covariance of the adaptive factor correction, so that the corrected filter gain is matched with the movement of the target, so that a better filter is obtained The experiment shows that the algorithm is better than the UKF. The advantages of the fusion UT transform and the EKF have the advantages of improving the tracking performance and the less operation time of the algorithm. A target tracking algorithm. (1) Unsensitive extension of Carl The Kalman filter tracking algorithm. UKF has better tracking performance than EKF in a non-linear tracking system, but the required calculation time is greater than E. Based on this reason, a target tracking method of fusion-insensitive transform (UT) and extended Kalman filter is proposed, which is mainly used to convert the state covariance of UKF and the multi-vector of the mutual covariance of the state and the measurement. a plurality of addition calculations to effectively reduce The operation time of the algorithm is as follows: the characteristics of the diversity Sigma particles transformed by the algorithm and the characteristics of the operation time of the EKF are fast, so that the better filtering and tracking precision is preserved, and the algorithm less computation time. (2) Adaptive non-sensitive extension of Carl In the process of the non-sensitive extended Kalman filter, the residual information is used to estimate and correct the two noise covariance in real time by means of exponential decay and forgetting factor, so as to realize the noise coordination. The experimental results show that the two algorithms have better tracking precision than UKF and have the same In order to improve the accuracy of the model probability estimation, a model-based probability model is proposed. An interactive multi-mode algorithm is used to calculate the state information after filtering. The calculation of the weighting factor (i.e., the probability of the model) mainly uses two types of information: the new interest and the model probability prediction value, and the method does not utilize the effective information of the current time state covariance to cause the opposite mode. based on this characteristic, the information of the state covariance is fused to obtain another weighting factor, and the model probability estimation value in the IMM algorithm is corrected by the weighting factor, that is, the weight The algorithm not only uses the prediction model probability factor but also uses the current state variance weighting factor, so it has more accuracy. The model selection probability is estimated by the experiment. The experiment verifies that the algorithm is more accurate than the IMM Finally, the paper sums up the work of the paper, and points out that the pape
【学位授予单位】：西安电子科技大学
【学位级别】：博士
【学位授予年份】：2014
【分类号】：TN953

【相似文献】