【作者】 刘娟丽；

【导师】 姬红兵；

【作者基本信息】 西安电子科技大学 ， 信号与信息处理， 2010， 硕士

【摘要】 被动多传感器的机动目标跟踪是目标跟踪领域的一个重要研究方面,受到越来越多国内外学者专家的关注。随着现代科学技术的日益发展,目标的机动性能不断提高,使得现代战争环境日益复杂,对目标跟踪提出越来越高的要求。因而,研究被动多传感器系统下的机动目标跟踪,对于提高我国防御系统的能力,具有十分重要的意义。针对机动目标的机动特性,本文重点研究了基于交互多模型(IMM)的机动目标跟踪算法。本文首先针对被动多传感器的非线性观测问题,研究了最近新提出的基于确定性采样的高斯滤波算法,详细阐述了利用拉格朗日乘数法得到更加逼近状态先验概率密度函数采样点的过程。利用IMM算法典型的模块化设计的优点,本文提出了一种基于交互多模型高斯滤波(IMMGF)的被动多传感器目标跟踪算法,该算法能够有效地提高机动目标跟踪的精度。其次,针对由于IMM算法采用标量权值更新模型概率而导致的目标的速度、加速度等估计不精确问题,研究了基于对角矩阵的交互多模型(DIMM)算法,该算法利用最优信息融合理论,将各模型状态估计的每一维向量元素(如：位置、速度、加速度等)实现优化融合,使得融合后状态的误差协方差矩阵最小。本文结合无迹卡尔曼滤波(UKF)算法,提出了DIMMUKF算法,该算法有效地提高了机动目标跟踪的精度,尤其对目标速度、加速度等的估计更加精确。再次,研究了基于模型切换时间的交互多模型(STC-IMM)算法,该算法在传感器的采样频率大于目标的机动频率的情况下,机动跟踪性能明显优于IMM算法。本文将STC-IMM算法与扩展卡尔曼滤波(EKF)算法相结合,应用于被动多传感器系统中,提出了一种新的STC-IMM-EKF算法,实现了只测角条件下对机动目标的跟踪。最后,针对机动目标跟踪中模型难以匹配的问题,本文提出了一种基于曲线模型自适应的机动目标跟踪算法。该算法将已有的曲线模型自适应跟踪算法作了一些改进,有效地避免了由于模型估计不精确导致的滤波器发散的现象。仿真实验表明,本文提出算法比现有算法取得更高精度的同时,扩大了其应用范围。更多还原

【Abstract】 Maneuvering target tracking based on multiple passive sensors is one of important aspects in the field of target tracking, which has been paid attention by domestic and foreign scholars and experts. With the development of modern science and technology and the unceasing enhancement of the target maneuver performance, modern war environment is increasingly complex, which put forward higher requirement to target tracking. Therefore, the research of the maneuvering target tracking based on multiple passive sensors is of great significance to enhancing the capacity of China’s defense system. Aiming at maneuvering targets maneuvering characteristic, this dissertation focuses on Interacting Multiple Model (IMM) algorithm.Firstly, aiming at the nonlinear observation problem of multiple passive sensors, this dissertation involves researches about the recently proposed Gaussian Filter algorithm based on deterministic sampling, and elaborates in detail on utilizing Lagrange multiplier method so as to attain sampling points which is close to the state a priori probability density function. Combined with IMM algorithm modular design, the Interacting Multiple Model Gaussian filter algorithm (IMMGF) based on multiple passive sensors is proposed, which can effectively improve the accuracy of maneuvering target tracking.Secondly, in view of the imprecise estimated problem of the target’s velocity and acceleration resulting from the scalar weight to update model in the IMM algorithm, this dissertation involves researches about diagonal interacting multiple model (DIMM) algorithm, which utilizes the optimal information fusion theory so as that each dimension element of the state estimation vector (such as:position, velocity, acceleration, etc.) achieves optimal integration and state fusion error covariance matrix is minimized. In combination with the Unscented Kalman filter (UKF) algorithm, the DIMMUKF algorithm is proposed, which effectively improved the accuracy of maneuvering target tracking, especially for the target speed, acceleration, etc.Thirdly, this dissertation involves researches about a new Interacting Multiple Model with Switch Time Conditions (STC-IMM). When the sampling rate of the underlying continuous process is high compared to the target dynamics, the maneuver tracking performance of STC-IMM algorithm is significantly superior to the IMM algorithm. The IMM algorithm in combination with Extended Kalman Filter (EKF) applied to the passive multi-sensor maneuvering target tracking, a novel STC-IMM-EKF algorithm is proposed, which can achieve maneuvering target tracking in multi-sensor bearing-only tracking.Finally, to solve the problem that the model for maneuvering target tracking is difficult to match, an adaptive tracking algorithm based on curve model for maneuvering target tracking is proposed. It makes some improvements to the present curve model adaptive tracking algorithm, and then effectively avoids the filter divergence resulting from the imprecise estimates to the model. Simulation results show that the proposed algorithm has higher precision and broader scope of application than the existing algorithm.更多还原