【作者】 王旭东；

【导师】 兰朝凤；

【作者基本信息】 哈尔滨理工大学 ， 电子信息（专业学位）， 2024， 硕士

【摘要】 随着自动化和智能化技术的快速发展,目标跟踪算法在众多领域中扮演着关键角色,例如国防安全、自动驾驶、人机交互和医疗影像分析等。高效准确的目标跟踪算法是实现这些应用的基础,因此,探索和改进目标跟踪算法具有重要的研究意义和应用价值。基于上述背景,本文深入研究目标跟踪中的单目标跟踪算法和多目标跟踪算法,旨在提高跟踪系统在各种环境中的准确性。首先,深入探讨了目标跟踪算法的相关理论,包括状态空间模型和滤波方法,详细讨论了各种运动模型及其在目标跟踪中的应用,以及贝叶斯滤波、卡尔曼滤波、扩展卡尔曼滤波、无迹卡尔曼滤波和粒子滤波的算法原理。其次,重点研究了单目标的方位跟踪算法,以修正极坐标系(Modified Polar-coordinates,MPC)为基础,通过仿真实验验证了修正极坐标系下扩展卡尔曼滤波器(Modified Polar-coordinates Extended Kalman Filter,MPEKF)在目标方位跟踪系统中的有效性和准确性,针对MPEKF的发散问题,研究了修正极坐标系下无迹卡尔曼滤波器(Modified Polar-coordinates Unscented Kalman Filter,MPUKF),仿真实验结果表明MPUKF在目标方位跟踪系统中的准确性和稳定性等整体性能上更占优势。本文基于MPUKF引入噪声自适应系数,提出一种修正极坐标系下自适应无迹卡尔曼滤波器(Modified Polar-coordinates Adaptive Unscented Kalman Filter,MPAUKF),以提高MPUKF在目标方位跟踪系统中的性能,仿真实验结果表明MPAUKF相比于MPUKF具有更快的收敛速度和更好的准确性。最后,针对多目标跟踪系统在高杂波环境下存在目标新生和消亡的情况,导致动态目标数不确定性的问题,研究基于随机有限集的多目标跟踪算法。本文基于传统的高斯混合势概率假设密度(Gaussian mixtures-Cardinalized Probability Hypothesis Density,GM-CPHD)滤波器,提出了一种基于多特征(Multi featured,MF)辅助的MFGM-CPHD滤波器。通过引入多普勒及幅度信息对目标量测状态进行扩展,利用多普勒及幅度信息的联合量测似然函数代替单一信息的量测似然函数,增强了目标与杂波的区分度。在模拟的高杂波和已知信噪比环境中进行仿真实验,探讨MFGM-CPHD算法在目标数量估计和跟踪精度等方面的性能。研究表明,MFGM-CPHD算法可以将目标与杂波有效的区分开来,提高了多目标跟踪的准确性。更多还原

【Abstract】 With the rapid development of automation and intelligent technology,target tracking algorithms play an increasingly critical role in many fields,such as national defense and security,autonomous driving,human-computer interaction,and medical image analysis.Efficient and accurate target tracking algorithms are the basis for realizing these applications.Therefore,exploring and improving target tracking algorithms has important research significance and application value.Based on the above background,this paper conducts an in-depth study of single target tracking algorithms and multi-target tracking algorithms in target tracking,aiming to improve the accuracy of the tracking system in various environments.First,the relevant theories of target tracking algorithms are discussed in depth,including state space models and filtering methods.Various motion models and their applications in target tracking are discussed in detail,especially Bayesian filtering,Kalman filtering,and extended Kalman.Principles of filtering,unscented Kalman filtering and particle filtering and their performance in dealing with target tracking problems.Secondly,the azimuth tracking algorithm of a single target is emphatically studied,based on the modified polar coordinate system,the azimuth tracking of a single target is carried out,the effectiveness and accuracy of the extended Kalman filter in the target azimuth tracking system under the modified polar coordinate system are verified by simulation experiments,and the unscented Kalman filter in the modified polar coordinate system is studied for the divergence problem of MPEKF,and the simulation results show that MPUKF has more advantages in the overall performance of the target azimuth tracking system,such as accuracy and stability.In this paper,based on the noise adaptation coefficient introduced by MPUKF,an adaptive unscented Kalman filter based on modified polar coordinate system is studied to improve the performance of MPUKF in target azimuth tracking system,and the simulation results show that MPAUKF has faster convergence speed and higher accuracy than MPUKFFinally,to address the uncertainty of multi-target tracking systems in high clutter environments and the number of dynamic targets,a passive multi-target tracking algorithm based on random finite sets is studied when considering the birth and death of targets.Based on the traditional GM-CPHD filter,this paper proposes a multi-feature assisted MFGM-CPHD filter.The target measurement status is expanded by introducing Doppler and amplitude information,and the joint measurement likelihood function of Doppler and amplitude information is used to replace the measurement likelihood function of single information to enhance the distinction between targets and clutter.Simulation experiments were conducted in a simulated high clutter and known signal-to-noise ratio environment to explore the performance of the MFGM-CPHD algorithm in target number estimation and tracking accuracy.Research shows that the MFGM-CPHD algorithm can effectively distinguish targets from clutter and improve the accuracy of multi-target tracking.更多还原