【作者】 黄剑；

【导师】 胡卫东；

【作者基本信息】 国防科学技术大学 ， 信息与通信工程， 2013， 博士

【摘要】 低轨空间目标监测是进行空间态势感知、碰撞预警的重要基础，是利用空间资源、实现在轨空间安全的基本保证。构建低轨空间监测系统，实现低轨道目标的编目和跟踪任务具有重要意义。其中，探测信息处理是监测系统的核心，针对探测信息处理的技术研究不仅能够有效提升当前系统性能，适应未来监测需求，同时也可以从理论上指导构建更加合理的监测系统。地基雷达作为低轨空间监测的主要设备，是获取低轨空间目标观测数据的主要来源。本文以低轨空间目标雷达监测系统，尤其是新的篱笆型监测雷达系统为研究背景，针对雷达监测系统构造过程中探测信息处理的关键技术和未来监测需求中具有普适性的前沿问题开展研究，主要的研究工作和取得的一些成果如下：首先，分析了低轨空间目标环境，按照对空间目标属性认知的层次综述了低轨空间目标监测技术的发展现状和趋势，并阐述了低轨空间目标雷达探测信息处理的难点问题。此外，还对低轨空间目标的轨道运动特性和雷达探测模型的交叉学科的基础知识进行了梳理。其次，研究了基于篱笆型雷达监测系统的弱小目标信号检测和参数估计问题。首先介绍了空间目标的雷达观测模型，分析了目标穿越波束屏障的稀疏特性，并通过轨道知识挖掘了篱笆型体制下目标的方位角和径向加速度的耦合关系，将原有的三维空间重构问题降维到二维空间，从而提升计算效率，改善重构精度。在此基础上，提出了用于弱小目标信号检测和参数估计的含轨道知识约束的稀疏重构方法。最后通过理论分析和大量仿真实验评估了正确稀疏重构概率、参数选择对估计精度影响，并对多目标的分辨能力进行了分析，结果表明了算法在低信噪比下优越的信号检测和参数估计性能和超分辨能力。然后，研究了双屏篱笆体制的空间监测雷达的屏间数据关联问题。首先对双屏监测雷达体制进行了阐述，以信息集合的方式描述了观测数据。在此基础上，分析了信息集合所要满足的轨道知识约束条件，基于假设检验的方式，提出了利用不同假设关联集合表现在关联量，即径向速度上的差异程度对数据进行关联的方法。之后，基于建立的简化场景从理论上分析了算法在三维空间上的目标分辨能力。最后利用NORAD发布的轨道根数及仿真数据验证了算法的有效性，特别是对Iridium33和Cosmos2251碰撞产生的高密度碎片云进行数据关联时，表现出了很高的正确关联率。之后，研究了含机动检测的多目标轨道相关技术。本文将机动检测和轨道相关作为两个不可分离的问题同时进行解决，更加符合实际情况。首先建立了一次切向速度方向变轨的目标运动模型和观测模型，然后提出了基于最大后验概率准则进行含机动检测的轨道相关原理和方法。为了计算最大后验概率，首先通过二阶锥规划算法求解了含约束的非线性最小二乘问题，实现了对机动参数的精确估计，其次通过JPDA算法计算最大后验概率，进行了机动事件的判别和轨道相关事件的确认。最后，通过理论推导和仿真实验对机动检测性能和关联性能进行了分析，验证了算法的有效性，同时对本文算法可推广应用的场景进行了阐释。最后，研究了低轨空间群目标的跟踪技术。首先介绍了群目标的运动模型和观测模型，提出了含群中心的最优贝叶斯跟踪滤波器。在此基础上，通过贝叶斯原理，将贝叶斯滤波器分解为目标状态预测模型、群中心预测模型，观测概率模型、群中心和目标间相互作用的马尔科夫随机场(MRF)模型，并进行了详细的说明和求解。特别是通过建立群中心与目标间相互作用的MRF模型，使得我们既能够描述群目标整体运动趋势，又可以提升个体轨迹的跟踪精度。然后通过MCMC-Particle粒子滤波算法对以上贝叶斯跟踪滤波器进行实现，并分析了跟踪性能。最后又引入群的分离与合并机制，使得算法具备对多个群进行灵活跟踪的能力，提升了算法的实际应用价值，大量仿真结果表明了算法在低数据率、高杂波环境下的优越性能。更多还原

【Abstract】 Low earth orbit (LEO) space surveillance systems are playing a crucial andfundamental role to support important functionalities of the space situational awarenessand orbital collision avoidance, which provide the guarantees for efficient utilization ofthe space resources and safety of the movements of the on-orbit spacecrafts. Theconstruction of an effective LEO space surveillance system that undertakes thecataloging and tracking tasks for LEO space objects is very important and challenging.Moreover, the information processing approach is a core component in any surveillancesystem. The investigation on information processing techniques will improve not onlyon the performances of the existing systems to satisfy the future needs of theestablishments of space surveillance systems, but also get a better understanding ontheoretical developments of more effective surveillance systems.Serving as a major instrument of LEO space surveillance systems, theground-based radar systems are the dominating sources where the measurement data ofLEO space objects can be obtained. Therefore, the radar systems of a LEO spacesurveillance system, especially an emerging fence-type surveillance radar system, havebeen taken as the fundamental infrastructure in this thesis. Based on such aninfrastructure, we investigate the key techniques of information processing to constructa radar surveillance system and the general problems with frontiers to satisfy the futuredemands of space surveillance systems. The main results and contributions aresummarized as follows.Firstly, the LEO space object environments are theoretically analyzed thatfollowed by the introductions of the development status and tendency of thesurveillance technologies of the LEO space objects based on the cognition progress forspace objects’ characteristics. The main challenging issues in the informationprocessing for exploring the LEO space objects by using a radar system are alsoelaborated. In addition, the basic interdisciplinary knowledge including the orbitalmovements and radar exploring technologies are also refined.Secondly, the methodologies of signal detection and parameter estimation for thesmall space objects based on a fence-type space surveillance radar system areinvestigated. Moreover, the observation model of space objects based on the fence-typeradar system is also introduced. The sparse characteristics of space objects crossing thebeam fence are analyzed. Furthermore, the reconstruction problem in originalthree-dimensional space can be reduced into two-dimensional space by carefullyanalyzing the relations between the acceleration and the directions of arrival for thecorresponding LEO space debris based on a fence-type surveillance system, which canachieve a higher efficiency and improve the reconstruction accuracy. According to these, a sparse reconstruction method involving the orbital knowledge constraint is proposedfor the signal detection and parameter estimation of the small objects. Finally, thecorrect sparse reconstruction probability and the estimation accuracy affected by thevalue of selected parameters are evaluated by theoretical analysis and numeroussimulations. In addition, the resolution capability for multiple objects is also analyzed.The results demonstrate the robustness of the approach in scenarios with a lowSignal-to-Noise Ratio (SNR) and the super-resolution properties.Subsequently, the data association problem for LEO space debris surveillancebased on a double fence radar system is also investigated. The surveillance mechanismsof a double fence radar system are elaborated that followed by the descriptions of theobservation data by using the information sets. Based on these constructions, weanalyze the set of orbital constraints on the LEO space debris in which the informationsets have to be satisfied. Moreover, combining with the hypothesis test methods, a noveldata association scheme is implemented by analyzing the discrepancy of the associationvariables, i.e. radial velocities, which are calculated according to the differenthypothetical associated sets. Furthermore, we also derive a theoretical analysis of theresolution performance in three-dimensional space for our proposed schemes. Thesuperiority and the effectiveness of our novel data association scheme are demonstratedby experimental results. The data used in our experiments is the LEO space debriscatalog produced by the North American Air Defense Command (NORAD) up to2009,especially for scenarios with high densities of LEO space debris, which were primarilyproduced by the collisions between Iridium33and Cosmos2251, which highly supportthe demonstration of the double fence space surveillance radar system.In this thesis, we also explore the orbit correlation approach including themaneuver detection problems. We integrate these two problems mentioned above intoone interrelated problem, and consider them simultaneously under a scenario wherespace objects only perform a single in-track orbital maneuver during the time intervalsbetween observations. More precisely, we mathematically formulate such an integratedproblem as the Maximum A-Posteriori Probability (MAP) estimation. To solve theMAP estimation, the maneuvering parameters are firstly estimated by optimally solvingthe constrained non-linear least squares iterative process based on a Second-order ConeProgramming (SOCP) algorithm. Subsequently, the corresponding posterior probabilityof an orbital maneuver and a joint association event can be approximately calculated bythe Joint Probabilistic Data Association (JPDA) algorithm. The desired solution hasbeen derived based on the MAP criterions. The performances and advantages of theproposed approaches have been shown by both theoretical analysis and simulationresults. We have to address that the proposed algorithms can be adapted and extended tomany different situations.Finally, the group tracking methods of LEO space objects are studied as well. In this chapter, we firstly introduce the orbital movement model and the observation modelof group objects and propose the optimal Bayesian tracking filtering involving groupcenter. Subsequently, due to the Bayesian theorem, the Bayesian tracking procedure canbe broke down into some detailed modules including the state transition model of spaceobjects, the state transition model of the group centers, the interaction Markov RandomField (MRF) model between group centers and individual trajectories, and the posteriordensity model of observations. We mainly focus on how to use the interaction MRFmodel between group centers and individual trajectories. It has been shown that we canobtain not only a more robust estimation of object numbers and improve the accuracy ofthe estimated corresponding individual trajectory, but also depict the evolution of thegroups under scenarios with the low object detection probabilities. MCMC-Particlealgorithm has been utilized to calculate the Bayesian integral and fulfill group tracking.Furthermore, the mechanism for group configuration inference has been incorporatedinto our approach that makes the operations of merge and split for groups much moresmart and efficient during the tracking process. We also show that the proposedalgorithm has significant impacts for the practical applications. Finally, we evaluate theperformances of our algorithms by the simulations of tracking multiple closely spacedorbital objects. The results verified the effectiveness of our proposed schemes for thescenarios with a low detection probability in a high dense clutter.更多还原